Skip to content
Sign in

flagon-io/g1t

public
0
1,211 linesCodeBlameRaw

Pick any line to see why it is the way it is: the commit, the pull request and issue it came from, and what the agent was thinking.

Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step1//! Scopes: what an access token may do on its owner's behalf.
2//!
3//! A personal access token, a workspace's token and an application signed
4//! in with OAuth each carry a set of scopes. A token reaches whatever the
5//! one it acts as can reach: a person's token, that person's workspaces and
6//! repositories; a workspace's token, that workspace. What a request may do
7//! is the intersection of two things: the role of whoever the token acts as
8//! (see [`crate::access`]) and the token's scopes.
9//!
10//! Each scope is a resource and a level, written `resource:level`, such as
11//! `issues:write`. A higher level of a resource includes the lower ones:
12//! `repo:admin` includes `repo:write`, which includes `repo:read`.
13//!
14//! This module is the one source of truth: the API (REST and MCP) and git
15//! enforce it, and identity stores it. `packages/contracts/src/scopes.ts`
16//! mirrors the table for the site; a test keeps the two the same.
17
18use serde::{Deserialize, Serialize};
19
20use crate::credentials::Decision;
21
22/// Something a token can be given access to.
23#[derive(Clone, Copy, Debug, PartialEq, Eq, Hash)]
24pub enum Resource {
25 Account,
API: notifications over REST and MCP, with notifications scopes26 Notifications,
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step27 Workspace,
Usage, Billing settings and prepaid AI credit; fixes from the UX audit28 Billing,
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step29 Repo,
30 Code,
Teams and CODEOWNERS, labels and milestones, dependency updates, the security suite, and a clearer top bar31 Security,
Packages, with a container registry on g1t.sh; workspaces deleted whole and kept 30 days; Members for every member32 Packages,
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step33 Issues,
34 PullRequests,
35 Agents,
36 Workflows,
Merge checks: statuses and check runs on every commit37 Checks,
Merge branch 'main' into worktree-agent-a69aeabc4b0deeb9738 Deployments,
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step39 Memory,
40 Access,
41 Webhooks,
42 Secrets,
Fast pages, required checks on the branch, self-hosted runners, honest incidents43 Runners,
Merge the AI Gateway: Anthropic's Messages API on a workspace's tokens44 Models,
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step45}
46
47impl Resource {
Merge checks: statuses and check runs on every commit48 pub const ALL: [Resource; 20] = [
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step49 Resource::Repo,
50 Resource::Code,
Teams and CODEOWNERS, labels and milestones, dependency updates, the security suite, and a clearer top bar51 Resource::Security,
Packages, with a container registry on g1t.sh; workspaces deleted whole and kept 30 days; Members for every member52 Resource::Packages,
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step53 Resource::Issues,
54 Resource::PullRequests,
55 Resource::Agents,
56 Resource::Workflows,
Merge checks: statuses and check runs on every commit57 Resource::Checks,
Merge branch 'main' into worktree-agent-a69aeabc4b0deeb9758 Resource::Deployments,
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step59 Resource::Memory,
60 Resource::Account,
API: notifications over REST and MCP, with notifications scopes61 Resource::Notifications,
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step62 Resource::Workspace,
Usage, Billing settings and prepaid AI credit; fixes from the UX audit63 Resource::Billing,
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step64 Resource::Access,
65 Resource::Webhooks,
66 Resource::Secrets,
Fast pages, required checks on the branch, self-hosted runners, honest incidents67 Resource::Runners,
Merge the AI Gateway: Anthropic's Messages API on a workspace's tokens68 Resource::Models,
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step69 ];
70
71 pub fn as_str(self) -> &'static str {
72 match self {
73 Resource::Account => "account",
API: notifications over REST and MCP, with notifications scopes74 Resource::Notifications => "notifications",
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step75 Resource::Workspace => "workspace",
Usage, Billing settings and prepaid AI credit; fixes from the UX audit76 Resource::Billing => "billing",
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step77 Resource::Repo => "repo",
78 Resource::Code => "code",
Teams and CODEOWNERS, labels and milestones, dependency updates, the security suite, and a clearer top bar79 Resource::Security => "security",
Packages, with a container registry on g1t.sh; workspaces deleted whole and kept 30 days; Members for every member80 Resource::Packages => "packages",
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step81 Resource::Issues => "issues",
82 Resource::PullRequests => "pull_requests",
83 Resource::Agents => "agents",
84 Resource::Workflows => "workflows",
Merge checks: statuses and check runs on every commit85 Resource::Checks => "checks",
Merge branch 'main' into worktree-agent-a69aeabc4b0deeb9786 Resource::Deployments => "deployments",
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step87 Resource::Memory => "memory",
88 Resource::Access => "access",
89 Resource::Webhooks => "webhooks",
90 Resource::Secrets => "secrets",
Fast pages, required checks on the branch, self-hosted runners, honest incidents91 Resource::Runners => "runners",
Merge the AI Gateway: Anthropic's Messages API on a workspace's tokens92 Resource::Models => "models",
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step93 }
94 }
95
96 /// Its name, for people.
97 pub fn label(self) -> &'static str {
98 match self {
99 Resource::Account => "Your account",
API: notifications over REST and MCP, with notifications scopes100 Resource::Notifications => "Notifications",
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step101 Resource::Workspace => "Workspaces",
Usage, Billing settings and prepaid AI credit; fixes from the UX audit102 Resource::Billing => "Billing",
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step103 Resource::Repo => "Repositories",
104 Resource::Code => "Code",
Teams and CODEOWNERS, labels and milestones, dependency updates, the security suite, and a clearer top bar105 Resource::Security => "Security",
Packages, with a container registry on g1t.sh; workspaces deleted whole and kept 30 days; Members for every member106 Resource::Packages => "Packages",
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step107 Resource::Issues => "Issues",
108 Resource::PullRequests => "Pull requests",
109 Resource::Agents => "g1t agents",
110 Resource::Workflows => "Workflows",
Merge checks: statuses and check runs on every commit111 Resource::Checks => "Checks and statuses",
Merge branch 'main' into worktree-agent-a69aeabc4b0deeb97112 Resource::Deployments => "Deployments",
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step113 Resource::Memory => "Memory and context",
114 Resource::Access => "Who has access",
115 Resource::Webhooks => "Webhooks",
116 Resource::Secrets => "Secrets and variables",
Fast pages, required checks on the branch, self-hosted runners, honest incidents117 Resource::Runners => "Self-hosted runners",
Merge the AI Gateway: Anthropic's Messages API on a workspace's tokens118 Resource::Models => "AI Gateway",
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step119 }
120 }
121}
122
123/// How much of a resource.
124#[derive(Clone, Copy, Debug, PartialEq, Eq, PartialOrd, Ord, Hash)]
125pub enum Level {
126 Read,
127 Write,
128 /// Starting g1t's agents, which spends the workspace's money.
129 Run,
Packages, with a container registry on g1t.sh; workspaces deleted whole and kept 30 days; Members for every member130 /// Deleting what cannot be brought back, such as a package's versions.
131 Delete,
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step132 Admin,
133}
134
135impl Level {
136 pub fn as_str(self) -> &'static str {
137 match self {
138 Level::Read => "read",
139 Level::Write => "write",
140 Level::Run => "run",
Packages, with a container registry on g1t.sh; workspaces deleted whole and kept 30 days; Members for every member141 Level::Delete => "delete",
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step142 Level::Admin => "admin",
143 }
144 }
145}
146
147/// One scope. Its text form, `resource:level`, is what tokens store, OAuth
148/// clients ask for, and errors name.
149#[derive(Clone, Copy, Debug, PartialEq, Eq, Hash)]
150pub enum Scope {
151 RepoRead,
152 RepoWrite,
153 RepoAdmin,
154 CodeRead,
155 CodeWrite,
Teams and CODEOWNERS, labels and milestones, dependency updates, the security suite, and a clearer top bar156 SecurityRead,
157 SecurityWrite,
Packages, with a container registry on g1t.sh; workspaces deleted whole and kept 30 days; Members for every member158 PackagesRead,
159 PackagesWrite,
160 PackagesDelete,
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step161 IssuesRead,
162 IssuesWrite,
163 PullRequestsRead,
164 PullRequestsWrite,
165 AgentsRun,
166 WorkflowsRead,
167 WorkflowsWrite,
Merge checks: statuses and check runs on every commit168 ChecksRead,
169 ChecksWrite,
Merge branch 'main' into worktree-agent-a69aeabc4b0deeb97170 DeploymentsRead,
171 DeploymentsWrite,
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step172 MemoryRead,
173 MemoryWrite,
174 AccountRead,
175 AccountWrite,
API: notifications over REST and MCP, with notifications scopes176 NotificationsRead,
177 NotificationsWrite,
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step178 WorkspaceRead,
179 WorkspaceAdmin,
Usage, Billing settings and prepaid AI credit; fixes from the UX audit180 BillingRead,
181 BillingWrite,
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step182 AccessRead,
183 AccessAdmin,
184 WebhooksRead,
185 WebhooksAdmin,
186 SecretsRead,
187 SecretsAdmin,
Fast pages, required checks on the branch, self-hosted runners, honest incidents188 RunnersRead,
189 RunnersAdmin,
Merge the AI Gateway: Anthropic's Messages API on a workspace's tokens190 ModelsRead,
191 ModelsWrite,
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step192}
193
194impl Scope {
195 /// Every scope, grouped by resource, least first.
Merge checks: statuses and check runs on every commit196 pub const ALL: [Scope; 41] = [
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step197 Scope::RepoRead,
198 Scope::RepoWrite,
199 Scope::RepoAdmin,
200 Scope::CodeRead,
201 Scope::CodeWrite,
Teams and CODEOWNERS, labels and milestones, dependency updates, the security suite, and a clearer top bar202 Scope::SecurityRead,
203 Scope::SecurityWrite,
Packages, with a container registry on g1t.sh; workspaces deleted whole and kept 30 days; Members for every member204 Scope::PackagesRead,
205 Scope::PackagesWrite,
206 Scope::PackagesDelete,
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step207 Scope::IssuesRead,
208 Scope::IssuesWrite,
209 Scope::PullRequestsRead,
210 Scope::PullRequestsWrite,
211 Scope::AgentsRun,
212 Scope::WorkflowsRead,
213 Scope::WorkflowsWrite,
Merge checks: statuses and check runs on every commit214 Scope::ChecksRead,
215 Scope::ChecksWrite,
Merge branch 'main' into worktree-agent-a69aeabc4b0deeb97216 Scope::DeploymentsRead,
217 Scope::DeploymentsWrite,
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step218 Scope::MemoryRead,
219 Scope::MemoryWrite,
220 Scope::AccountRead,
221 Scope::AccountWrite,
API: notifications over REST and MCP, with notifications scopes222 Scope::NotificationsRead,
223 Scope::NotificationsWrite,
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step224 Scope::WorkspaceRead,
225 Scope::WorkspaceAdmin,
Usage, Billing settings and prepaid AI credit; fixes from the UX audit226 Scope::BillingRead,
227 Scope::BillingWrite,
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step228 Scope::AccessRead,
229 Scope::AccessAdmin,
230 Scope::WebhooksRead,
231 Scope::WebhooksAdmin,
232 Scope::SecretsRead,
233 Scope::SecretsAdmin,
Fast pages, required checks on the branch, self-hosted runners, honest incidents234 Scope::RunnersRead,
235 Scope::RunnersAdmin,
Merge the AI Gateway: Anthropic's Messages API on a workspace's tokens236 Scope::ModelsRead,
237 Scope::ModelsWrite,
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step238 ];
239
240 pub fn as_str(self) -> &'static str {
241 match self {
242 Scope::RepoRead => "repo:read",
243 Scope::RepoWrite => "repo:write",
244 Scope::RepoAdmin => "repo:admin",
245 Scope::CodeRead => "code:read",
246 Scope::CodeWrite => "code:write",
Teams and CODEOWNERS, labels and milestones, dependency updates, the security suite, and a clearer top bar247 Scope::SecurityRead => "security:read",
248 Scope::SecurityWrite => "security:write",
Packages, with a container registry on g1t.sh; workspaces deleted whole and kept 30 days; Members for every member249 Scope::PackagesRead => "packages:read",
250 Scope::PackagesWrite => "packages:write",
251 Scope::PackagesDelete => "packages:delete",
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step252 Scope::IssuesRead => "issues:read",
253 Scope::IssuesWrite => "issues:write",
254 Scope::PullRequestsRead => "pull_requests:read",
255 Scope::PullRequestsWrite => "pull_requests:write",
256 Scope::AgentsRun => "agents:run",
257 Scope::WorkflowsRead => "workflows:read",
258 Scope::WorkflowsWrite => "workflows:write",
Merge checks: statuses and check runs on every commit259 Scope::ChecksRead => "checks:read",
260 Scope::ChecksWrite => "checks:write",
Merge branch 'main' into worktree-agent-a69aeabc4b0deeb97261 Scope::DeploymentsRead => "deployments:read",
262 Scope::DeploymentsWrite => "deployments:write",
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step263 Scope::MemoryRead => "memory:read",
264 Scope::MemoryWrite => "memory:write",
265 Scope::AccountRead => "account:read",
266 Scope::AccountWrite => "account:write",
API: notifications over REST and MCP, with notifications scopes267 Scope::NotificationsRead => "notifications:read",
268 Scope::NotificationsWrite => "notifications:write",
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step269 Scope::WorkspaceRead => "workspace:read",
270 Scope::WorkspaceAdmin => "workspace:admin",
Usage, Billing settings and prepaid AI credit; fixes from the UX audit271 Scope::BillingRead => "billing:read",
272 Scope::BillingWrite => "billing:write",
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step273 Scope::AccessRead => "access:read",
274 Scope::AccessAdmin => "access:admin",
275 Scope::WebhooksRead => "webhooks:read",
276 Scope::WebhooksAdmin => "webhooks:admin",
277 Scope::SecretsRead => "secrets:read",
278 Scope::SecretsAdmin => "secrets:admin",
Fast pages, required checks on the branch, self-hosted runners, honest incidents279 Scope::RunnersRead => "runners:read",
280 Scope::RunnersAdmin => "runners:admin",
Merge the AI Gateway: Anthropic's Messages API on a workspace's tokens281 Scope::ModelsRead => "models:read",
282 Scope::ModelsWrite => "models:write",
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step283 }
284 }
285
286 pub fn parse(text: &str) -> Option<Scope> {
287 let text = text.trim().to_ascii_lowercase();
288 Scope::ALL.into_iter().find(|scope| scope.as_str() == text)
289 }
290
291 pub fn resource(self) -> Resource {
292 let name = self.as_str().split_once(':').map_or("", |(resource, _)| resource);
293 Resource::ALL
294 .into_iter()
295 .find(|resource| resource.as_str() == name)
296 .unwrap_or(Resource::Account)
297 }
298
299 pub fn level(self) -> Level {
300 match self.as_str().rsplit_once(':').map_or("", |(_, level)| level) {
301 "write" => Level::Write,
302 "run" => Level::Run,
Packages, with a container registry on g1t.sh; workspaces deleted whole and kept 30 days; Members for every member303 "delete" => Level::Delete,
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step304 "admin" => Level::Admin,
305 _ => Level::Read,
306 }
307 }
308
309 /// Whether holding `self` gives `other`: the same resource, at the same
310 /// level or a lower one.
311 pub fn includes(self, other: Scope) -> bool {
312 self.resource() == other.resource() && self.level() >= other.level()
313 }
314
315 /// Changes that are hard or impossible to undo, or that decide who can
316 /// reach what. Shown behind a warning wherever scopes are chosen.
317 pub fn dangerous(self) -> bool {
Packages, with a container registry on g1t.sh; workspaces deleted whole and kept 30 days; Members for every member318 matches!(self.level(), Level::Admin | Level::Delete)
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step319 }
320
321 /// What it lets a token do, in plain words.
322 pub fn describe(self) -> &'static str {
323 match self {
Merge branch 'main' into worktree-agent-a69aeabc4b0deeb97324 Scope::RepoRead => "See repositories, their settings, labels, timelines, releases, languages, contributors and security alerts, and search",
325 Scope::RepoWrite => "Create repositories, rename branches, change how pull requests merge and publish releases",
Merge rulesets: branch and tag rules, agent-first, enforced on push and merge326 Scope::RepoAdmin => "Rename, archive, transfer, delete or change who can see a repository, change its rulesets, and dismiss security alerts",
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step327 Scope::CodeRead => "Clone and fetch private repositories with git",
328 Scope::CodeWrite => "Push commits with git",
Teams and CODEOWNERS, labels and milestones, dependency updates, the security suite, and a clearer top bar329 Scope::SecurityRead => "See secret scanning, code scanning and vulnerability alerts, custom patterns, the dependency graph and SBOM, and security settings",
330 Scope::SecurityWrite => "Dismiss and reopen alerts, bypass push protection, review bypass requests, manage custom patterns, upload SARIF and change security settings",
Packages, with a container registry on g1t.sh; workspaces deleted whole and kept 30 days; Members for every member331 Scope::PackagesRead => "Pull container images and install private packages",
332 Scope::PackagesWrite => "Push container images and publish packages",
333 Scope::PackagesDelete => "Delete packages and their versions",
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step334 Scope::IssuesRead => "Read issues, comments and plans",
335 Scope::IssuesWrite => "Open, edit, close and comment on issues",
336 Scope::PullRequestsRead => "Read pull requests, their changes, sessions and merge queues",
337 Scope::PullRequestsWrite => "Open, review, close and merge pull requests",
338 Scope::AgentsRun => "Put g1t agents to work and message them, which uses the workspace's money",
339 Scope::WorkflowsRead => "Read workflows, runs and logs",
340 Scope::WorkflowsWrite => "Run, cancel, rerun and turn workflows on or off",
Merge checks: statuses and check runs on every commit341 Scope::ChecksRead => "Read commits' statuses, check runs, check suites and annotations",
342 Scope::ChecksWrite => "Report statuses and check runs on commits, and ask for checks to run again",
Merge branch 'main' into worktree-agent-a69aeabc4b0deeb97343 Scope::DeploymentsRead => "See deployments, their statuses and environments",
344 Scope::DeploymentsWrite => "Report deployments and their statuses, from any CI",
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step345 Scope::MemoryRead => "Recall memory and search the workspace's context",
346 Scope::MemoryWrite => "Save memory for the next agent",
Merge branch 'main' into worktree-agent-a69aeabc4b0deeb97347 Scope::AccountRead => "Read your email addresses, invites, invitations, pinned projects and stars",
348 Scope::AccountWrite => "Change your email addresses, make invites, answer invitations, pin projects and star repositories",
API: notifications over REST and MCP, with notifications scopes349 Scope::NotificationsRead => "See your inbox, its threads, and what you subscribe to and watch",
350 Scope::NotificationsWrite => "Mark notifications read, done, saved or snoozed, subscribe to threads and watch repositories",
Merge rulesets: branch and tag rules, agent-first, enforced on push and merge351 Scope::WorkspaceRead => "Read workspace settings, invites, integrations, model routes, teams and rulesets",
352 Scope::WorkspaceAdmin => "Create and delete workspaces, invite members, connect integrations, create, change and delete teams, and change the workspace's rulesets",
Usage, Billing settings and prepaid AI credit; fixes from the UX audit353 Scope::BillingRead => "See a workspace's usage, budget, AI credit and invoices",
354 Scope::BillingWrite => "Change a workspace's budget and buy AI credit",
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step355 Scope::AccessRead => "See who has access to repositories",
Teams and CODEOWNERS, labels and milestones, dependency updates, the security suite, and a clearer top bar356 Scope::AccessAdmin => "Give and take away access to repositories, a team's included",
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step357 Scope::WebhooksRead => "See webhooks and their deliveries",
358 Scope::WebhooksAdmin => "Create, change and delete webhooks",
359 Scope::SecretsRead => "List secrets (never their values) and read variables",
360 Scope::SecretsAdmin => "Set and delete secrets and variables",
Fast pages, required checks on the branch, self-hosted runners, honest incidents361 Scope::RunnersRead => "See self-hosted runners, their groups and where agents run",
362 Scope::RunnersAdmin => "Register and remove self-hosted runners, change their groups and settings",
Merge the AI Gateway: Anthropic's Messages API on a workspace's tokens363 Scope::ModelsRead => "See the workspace's AI Gateway requests: their models, tokens, cost and status",
364 Scope::ModelsWrite => "Send model requests through the AI Gateway, which uses the workspace's AI credit",
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step365 }
366 }
367}
368
369impl Serialize for Scope {
370 fn serialize<S: serde::Serializer>(&self, serializer: S) -> Result<S::Ok, S::Error> {
371 serializer.serialize_str(self.as_str())
372 }
373}
374
375impl<'de> Deserialize<'de> for Scope {
376 fn deserialize<D: serde::Deserializer<'de>>(deserializer: D) -> Result<Self, D::Error> {
377 let text = String::deserialize(deserializer)?;
378 Scope::parse(&text).ok_or_else(|| serde::de::Error::custom(format!("unknown scope {text}")))
379 }
380}
381
382/// Scopes as written in a token's row or an OAuth request: separated by
383/// spaces or commas. Unknown names are left out, so a client asking for a
384/// scope from a newer version gets the rest.
385pub fn parse_scopes(text: &str) -> Vec<Scope> {
386 let mut scopes: Vec<Scope> = text
387 .split(|c: char| c.is_whitespace() || c == ',')
388 .filter_map(Scope::parse)
389 .collect();
390 normalize(&mut scopes);
391 scopes
392}
393
394/// In table order, without repeats.
395pub fn normalize(scopes: &mut Vec<Scope>) {
396 let given = std::mem::take(scopes);
397 scopes.extend(Scope::ALL.into_iter().filter(|scope| given.contains(scope)));
398}
399
400/// Space-separated, as stored and as OAuth writes them.
401pub fn scopes_text(scopes: &[Scope]) -> String {
402 scopes.iter().map(|scope| scope.as_str()).collect::<Vec<_>>().join(" ")
403}
404
405/// What a token stores for full access, which is not a scope a client can
406/// ask for by name.
407pub const FULL_ACCESS: &str = "*";
408
409/// Starting points for choosing scopes.
410#[derive(Clone, Copy, Debug, PartialEq, Eq)]
411pub enum Preset {
412 ReadOnly,
413 Agent,
414 Ci,
415 Full,
416}
417
418impl Preset {
419 pub const ALL: [Preset; 4] = [Preset::ReadOnly, Preset::Agent, Preset::Ci, Preset::Full];
420
421 pub fn as_str(self) -> &'static str {
422 match self {
423 Preset::ReadOnly => "read_only",
424 Preset::Agent => "agent",
425 Preset::Ci => "ci",
426 Preset::Full => "full",
427 }
428 }
429
430 pub fn label(self) -> &'static str {
431 match self {
432 Preset::ReadOnly => "Read only",
433 Preset::Agent => "Agent",
434 Preset::Ci => "CI",
435 Preset::Full => "Full access",
436 }
437 }
438
439 /// Its scopes; `None` for full access.
440 pub fn scopes(self) -> Option<Vec<Scope>> {
441 let reads = || Scope::ALL.into_iter().filter(|scope| scope.level() == Level::Read);
442 match self {
443 Preset::ReadOnly => Some(reads().collect()),
444 Preset::Agent => {
Fast pages, required checks on the branch, self-hosted runners, honest incidents445 // Not the machines work runs on: an agent has no business
446 // knowing a workspace's own runners.
447 let mut scopes: Vec<Scope> = reads().filter(|scope| scope.resource() != Resource::Runners).collect();
API: notifications over REST and MCP, with notifications scopes448 // And answering what needs the person it works for: marking
449 // it done, subscribing, watching.
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step450 scopes.extend([
451 Scope::CodeWrite,
452 Scope::IssuesWrite,
453 Scope::PullRequestsWrite,
454 Scope::AgentsRun,
455 Scope::MemoryWrite,
API: notifications over REST and MCP, with notifications scopes456 Scope::NotificationsWrite,
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step457 ]);
458 normalize(&mut scopes);
459 Some(scopes)
460 }
461 Preset::Ci => Some(vec![
462 Scope::RepoRead,
463 Scope::CodeRead,
464 Scope::CodeWrite,
Packages, with a container registry on g1t.sh; workspaces deleted whole and kept 30 days; Members for every member465 Scope::PackagesRead,
466 Scope::PackagesWrite,
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step467 Scope::WorkflowsRead,
468 Scope::WorkflowsWrite,
Merge checks: statuses and check runs on every commit469 Scope::ChecksRead,
470 Scope::ChecksWrite,
Merge branch 'main' into worktree-agent-a69aeabc4b0deeb97471 Scope::DeploymentsRead,
472 Scope::DeploymentsWrite,
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step473 ]),
474 Preset::Full => None,
475 }
476 }
477}
478
479/// What an OAuth client gets when it asks for nothing in particular: the
480/// agent preset. Never an admin scope.
481pub fn oauth_default() -> Vec<Scope> {
482 Preset::Agent.scopes().unwrap_or_default()
483}
484
485/// Set on a [`crate::User`] resolved from an access token: what the token
486/// may do. Absent on a signed-in session, which may do whatever its person
487/// can.
488#[derive(Clone, Debug, Default, PartialEq, Eq, Serialize, Deserialize)]
489pub struct TokenAccess {
490 /// The token's id, as audit entries and errors name it.
491 #[serde(default)]
492 pub token_id: String,
493 /// Its scopes, as `resource:level`. Absent: full access, everything the
494 /// person (or workspace) can do.
495 #[serde(default, skip_serializing_if = "Option::is_none")]
496 pub scopes: Option<Vec<String>>,
497 /// Made before tokens had scopes: full access until someone narrows it.
498 #[serde(default, skip_serializing_if = "std::ops::Not::not")]
499 pub legacy: bool,
Merge the AI Gateway: Anthropic's Messages API on a workspace's tokens500 /// The token's name, as its owner gave it, so a log can say which
501 /// token made a request. Absent where whoever resolved it did not say.
502 #[serde(default, skip_serializing_if = "Option::is_none")]
503 pub name: Option<String>,
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step504}
505
506impl TokenAccess {
507 /// Full access to everything: the access tokens made before scopes had.
508 pub fn full() -> Self {
509 TokenAccess::default()
510 }
511
512 pub fn is_full(&self) -> bool {
513 self.scopes.is_none()
514 }
515
516 /// The scopes it holds, or `None` for full access.
517 pub fn granted(&self) -> Option<Vec<Scope>> {
518 self.scopes
519 .as_ref()
520 .map(|scopes| scopes.iter().filter_map(|scope| Scope::parse(scope)).collect())
521 }
522
523 pub fn allows(&self, needed: Scope) -> bool {
524 match self.granted() {
525 None => true,
526 Some(granted) => granted.iter().any(|held| held.includes(needed)),
527 }
528 }
529}
530
531/// Every operation of the API and MCP server, with the scope it needs. An
532/// operation in [`NO_SCOPE`] needs none. The API checks that every one of
533/// its operations is in exactly one of the two.
534pub const OPERATIONS: &[(&str, Scope)] = &[
535 // Your account.
536 ("list_emails", Scope::AccountRead),
537 ("add_email", Scope::AccountWrite),
538 ("remove_email", Scope::AccountWrite),
539 ("update_email_settings", Scope::AccountWrite),
540 ("list_invites", Scope::AccountRead),
541 ("create_invite", Scope::AccountWrite),
542 ("revoke_invite", Scope::AccountWrite),
543 ("list_my_repo_invitations", Scope::AccountRead),
544 ("accept_repo_invitation", Scope::AccountWrite),
545 ("decline_repo_invitation", Scope::AccountWrite),
API: pinned projects over REST and MCP546 // Your pinned projects: a preference of your account.
547 ("list_pinned_projects", Scope::AccountRead),
548 ("pin_project", Scope::AccountWrite),
Merge branch 'main' into worktree-agent-a69aeabc4b0deeb97549 // Your stars: a preference of your account.
550 ("list_starred", Scope::AccountRead),
551 ("check_starred", Scope::AccountRead),
552 ("star_repo", Scope::AccountWrite),
553 ("unstar_repo", Scope::AccountWrite),
API: pinned projects over REST and MCP554 ("unpin_project", Scope::AccountWrite),
555 ("reorder_pinned_projects", Scope::AccountWrite),
API: notifications over REST and MCP, with notifications scopes556 // Your inbox: notifications, subscriptions and watching.
557 ("list_notifications", Scope::NotificationsRead),
558 ("get_notification_thread", Scope::NotificationsRead),
559 ("get_thread_subscription", Scope::NotificationsRead),
560 ("get_repo_subscription", Scope::NotificationsRead),
561 ("list_watched_repos", Scope::NotificationsRead),
562 ("mark_notifications_read", Scope::NotificationsWrite),
563 ("mark_thread_read", Scope::NotificationsWrite),
564 ("mark_thread_done", Scope::NotificationsWrite),
565 ("save_thread", Scope::NotificationsWrite),
566 ("snooze_thread", Scope::NotificationsWrite),
567 ("set_thread_subscription", Scope::NotificationsWrite),
568 ("delete_thread_subscription", Scope::NotificationsWrite),
569 ("set_repo_subscription", Scope::NotificationsWrite),
570 ("delete_repo_subscription", Scope::NotificationsWrite),
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step571 // Workspaces, their invites and integrations.
572 ("create_workspace", Scope::WorkspaceAdmin),
573 ("delete_workspace", Scope::WorkspaceAdmin),
Merge branch 'worktree-agent-ad7c6d88d93adc817'574 ("get_workspace", Scope::WorkspaceRead),
Git storage hardened, pages in tens of milliseconds, honest security alerts, and costs reconciled daily575 ("update_workspace", Scope::WorkspaceAdmin),
Membership as GitHub has it: owners, roles, member privileges, 2FA576 // Its members, and who owns it.
577 ("list_members", Scope::WorkspaceRead),
578 ("update_member", Scope::WorkspaceAdmin),
579 ("remove_member", Scope::WorkspaceAdmin),
580 ("transfer_ownership", Scope::WorkspaceAdmin),
581 ("leave_workspace", Scope::AccountWrite),
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step582 ("list_workspace_invites", Scope::WorkspaceRead),
583 ("invite_member", Scope::WorkspaceAdmin),
584 ("revoke_workspace_invite", Scope::WorkspaceAdmin),
585 ("list_integrations", Scope::WorkspaceRead),
586 ("connect_integration", Scope::WorkspaceAdmin),
AI Gateway: OpenAI's format, open models, and your own providers587 ("update_integration", Scope::WorkspaceAdmin),
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step588 ("disconnect_integration", Scope::WorkspaceAdmin),
589 ("test_integration", Scope::WorkspaceAdmin),
590 ("get_model_routes", Scope::WorkspaceRead),
591 ("set_model_routes", Scope::WorkspaceAdmin),
Teams and CODEOWNERS, labels and milestones, dependency updates, the security suite, and a clearer top bar592 // Teams: reading them, and managing them. A team's role on a
593 // repository is who has access.
594 ("list_teams", Scope::WorkspaceRead),
595 ("get_team", Scope::WorkspaceRead),
596 ("list_team_members", Scope::WorkspaceRead),
597 ("list_child_teams", Scope::WorkspaceRead),
598 ("list_team_repos", Scope::WorkspaceRead),
599 ("list_user_teams", Scope::WorkspaceRead),
600 ("create_team", Scope::WorkspaceAdmin),
Merge rulesets: branch and tag rules, agent-first, enforced on push and merge601 ("list_workspace_rulesets", Scope::WorkspaceRead),
602 ("get_workspace_ruleset", Scope::WorkspaceRead),
603 ("list_workspace_rule_evaluations", Scope::WorkspaceRead),
604 ("create_workspace_ruleset", Scope::WorkspaceAdmin),
605 ("update_workspace_ruleset", Scope::WorkspaceAdmin),
606 ("delete_workspace_ruleset", Scope::WorkspaceAdmin),
Teams and CODEOWNERS, labels and milestones, dependency updates, the security suite, and a clearer top bar607 ("update_team", Scope::WorkspaceAdmin),
608 ("delete_team", Scope::WorkspaceAdmin),
609 ("set_team_member", Scope::WorkspaceAdmin),
610 ("remove_team_member", Scope::WorkspaceAdmin),
611 ("set_team_review_assignment", Scope::WorkspaceAdmin),
Usage, Billing settings and prepaid AI credit; fixes from the UX audit612 // A workspace's billing: usage, budget, AI credit and invoices.
613 ("get_usage", Scope::BillingRead),
614 ("get_budget", Scope::BillingRead),
615 ("get_ai_credit", Scope::BillingRead),
616 ("list_invoices", Scope::BillingRead),
617 ("get_billing_details", Scope::BillingRead),
618 ("set_budget", Scope::BillingWrite),
619 ("buy_ai_credit", Scope::BillingWrite),
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step620 // Repositories.
621 ("list_repos", Scope::RepoRead),
622 ("get_repo", Scope::RepoRead),
Merge branch 'main' into worktree-agent-a69aeabc4b0deeb97623 // Projects follow their repositories.
624 ("list_projects", Scope::RepoRead),
625 ("get_project", Scope::RepoRead),
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step626 ("search", Scope::RepoRead),
627 ("list_events", Scope::RepoRead),
Merge branch 'main' into worktree-agent-a69aeabc4b0deeb97628 // What the default branch says about a repository, who starred it, and
629 // its releases.
630 ("get_languages", Scope::RepoRead),
631 ("list_contributors", Scope::RepoRead),
632 ("get_license", Scope::RepoRead),
633 ("list_stargazers", Scope::RepoRead),
634 ("list_releases", Scope::RepoRead),
635 ("get_latest_release", Scope::RepoRead),
636 ("get_release_by_tag", Scope::RepoRead),
637 ("get_release", Scope::RepoRead),
638 ("create_release", Scope::RepoWrite),
639 ("update_release", Scope::RepoWrite),
640 ("delete_release", Scope::RepoWrite),
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step641 ("list_labels", Scope::RepoRead),
Teams and CODEOWNERS, labels and milestones, dependency updates, the security suite, and a clearer top bar642 ("list_milestones", Scope::RepoRead),
643 ("get_milestone", Scope::RepoRead),
644 ("create_label", Scope::IssuesWrite),
645 ("update_label", Scope::IssuesWrite),
646 ("delete_label", Scope::IssuesWrite),
647 ("add_default_labels", Scope::IssuesWrite),
648 ("create_milestone", Scope::IssuesWrite),
649 ("update_milestone", Scope::IssuesWrite),
650 ("delete_milestone", Scope::IssuesWrite),
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step651 ("get_repo_settings", Scope::RepoRead),
Fast pages, required checks on the branch, self-hosted runners, honest incidents652 ("list_check_names", Scope::RepoRead),
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step653 ("list_deleted_repos", Scope::RepoRead),
Git storage hardened, pages in tens of milliseconds, honest security alerts, and costs reconciled daily654 ("list_security_alerts", Scope::RepoRead),
Teams and CODEOWNERS, labels and milestones, dependency updates, the security suite, and a clearer top bar655 ("get_codeowners_errors", Scope::RepoRead),
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step656 ("create_repo", Scope::RepoWrite),
657 ("update_repo", Scope::RepoWrite),
Merge branch 'main' into worktree-agent-a69aeabc4b0deeb97658 ("update_project", Scope::RepoWrite),
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step659 ("update_repo_settings", Scope::RepoWrite),
Merge rulesets: branch and tag rules, agent-first, enforced on push and merge660 // Rulesets: reading them is reading the repository; changing them
661 // changes what everyone, agents included, may do, so it is admin.
662 ("list_repo_rulesets", Scope::RepoRead),
663 ("get_repo_ruleset", Scope::RepoRead),
664 ("get_branch_rules", Scope::RepoRead),
665 ("list_rule_evaluations", Scope::RepoRead),
666 ("create_repo_ruleset", Scope::RepoAdmin),
667 ("update_repo_ruleset", Scope::RepoAdmin),
668 ("delete_repo_ruleset", Scope::RepoAdmin),
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step669 ("rename_branch", Scope::RepoWrite),
670 ("rename_repo", Scope::RepoAdmin),
671 ("transfer_repo", Scope::RepoAdmin),
672 ("archive_repo", Scope::RepoAdmin),
673 ("unarchive_repo", Scope::RepoAdmin),
674 ("set_repo_visibility", Scope::RepoAdmin),
675 ("delete_repo", Scope::RepoAdmin),
676 ("restore_repo", Scope::RepoAdmin),
677 ("purge_repo", Scope::RepoAdmin),
Git storage hardened, pages in tens of milliseconds, honest security alerts, and costs reconciled daily678 // A dismissed secret is let through push protection.
679 ("dismiss_security_alert", Scope::RepoAdmin),
680 ("reopen_security_alert", Scope::RepoAdmin),
Teams and CODEOWNERS, labels and milestones, dependency updates, the security suite, and a clearer top bar681 // The security suite: alerts, push protection, patterns, code
682 // scanning, the supply chain and settings.
683 ("list_secret_scanning_alerts", Scope::SecurityRead),
684 ("get_secret_scanning_alert", Scope::SecurityRead),
685 ("list_secret_scanning_locations", Scope::SecurityRead),
686 ("list_bypass_requests", Scope::SecurityRead),
687 ("list_custom_patterns", Scope::SecurityRead),
688 ("list_code_scanning_alerts", Scope::SecurityRead),
689 ("get_code_scanning_alert", Scope::SecurityRead),
690 ("list_code_scanning_analyses", Scope::SecurityRead),
691 ("get_sarif_upload", Scope::SecurityRead),
692 ("list_vulnerability_alerts", Scope::SecurityRead),
693 ("get_vulnerability_alert", Scope::SecurityRead),
694 ("get_dependency_graph", Scope::SecurityRead),
695 ("get_sbom", Scope::SecurityRead),
696 ("compare_dependencies", Scope::SecurityRead),
697 ("get_security_settings", Scope::SecurityRead),
698 ("get_workspace_security_settings", Scope::SecurityRead),
699 ("get_security_overview", Scope::SecurityRead),
700 ("update_secret_scanning_alert", Scope::SecurityWrite),
701 ("bypass_push_protection", Scope::SecurityWrite),
702 ("check_secret_validity", Scope::SecurityWrite),
703 ("review_bypass_request", Scope::SecurityWrite),
704 ("create_custom_pattern", Scope::SecurityWrite),
705 ("update_custom_pattern", Scope::SecurityWrite),
706 ("delete_custom_pattern", Scope::SecurityWrite),
707 ("dry_run_custom_pattern", Scope::SecurityWrite),
708 ("update_code_scanning_alert", Scope::SecurityWrite),
709 ("upload_sarif", Scope::SecurityWrite),
710 ("update_vulnerability_alert", Scope::SecurityWrite),
711 ("fix_security_alert", Scope::SecurityWrite),
712 ("update_security_settings", Scope::SecurityWrite),
713 ("update_workspace_security_settings", Scope::SecurityWrite),
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step714 // Issues and plans.
715 ("list_issues", Scope::IssuesRead),
716 ("get_issue", Scope::IssuesRead),
717 ("get_plan", Scope::IssuesRead),
718 ("create_issue", Scope::IssuesWrite),
719 ("update_issue", Scope::IssuesWrite),
Teams and CODEOWNERS, labels and milestones, dependency updates, the security suite, and a clearer top bar720 ("list_issue_labels", Scope::IssuesRead),
721 ("add_issue_labels", Scope::IssuesWrite),
722 ("set_issue_labels", Scope::IssuesWrite),
723 ("remove_issue_labels", Scope::IssuesWrite),
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step724 ("close_issue", Scope::IssuesWrite),
725 ("reopen_issue", Scope::IssuesWrite),
726 ("add_comment", Scope::IssuesWrite),
727 ("import_issue", Scope::IssuesWrite),
728 ("apply_plan", Scope::IssuesWrite),
729 // Pull requests.
730 ("list_pull_requests", Scope::PullRequestsRead),
731 ("get_pull_request", Scope::PullRequestsRead),
732 ("get_pull_request_changes", Scope::PullRequestsRead),
733 ("read_session", Scope::PullRequestsRead),
734 ("get_merge_queue", Scope::PullRequestsRead),
735 ("create_pull_request", Scope::PullRequestsWrite),
Teams and CODEOWNERS, labels and milestones, dependency updates, the security suite, and a clearer top bar736 ("update_pull_request", Scope::PullRequestsWrite),
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step737 ("record_session", Scope::PullRequestsWrite),
738 ("mark_pull_request_ready", Scope::PullRequestsWrite),
739 ("close_pull_request", Scope::PullRequestsWrite),
740 ("review_pull_request", Scope::PullRequestsWrite),
741 ("merge_pull_request", Scope::PullRequestsWrite),
Teams and CODEOWNERS, labels and milestones, dependency updates, the security suite, and a clearer top bar742 ("request_reviewers", Scope::PullRequestsWrite),
743 ("remove_requested_reviewers", Scope::PullRequestsWrite),
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step744 // g1t's agents.
745 ("assign_issue", Scope::AgentsRun),
746 ("delegate", Scope::AgentsRun),
747 ("plan_work", Scope::AgentsRun),
748 ("message_agent", Scope::AgentsRun),
749 ("answer_message", Scope::AgentsRun),
750 ("take_messages", Scope::AgentsRun),
751 // Workflows.
752 ("list_workflows", Scope::WorkflowsRead),
753 ("list_workflow_runs", Scope::WorkflowsRead),
754 ("get_workflow_run", Scope::WorkflowsRead),
755 ("get_job_logs", Scope::WorkflowsRead),
756 ("dispatch_workflow", Scope::WorkflowsWrite),
757 ("cancel_workflow_run", Scope::WorkflowsWrite),
758 ("rerun_workflow_run", Scope::WorkflowsWrite),
759 ("update_workflow", Scope::WorkflowsWrite),
Actions: OIDC tokens, the toolkit's cache and artifact services, and artifacts in R2760 ("list_artifacts", Scope::WorkflowsRead),
761 ("list_workflow_run_artifacts", Scope::WorkflowsRead),
762 ("get_artifact", Scope::WorkflowsRead),
763 ("download_artifact", Scope::WorkflowsRead),
764 ("get_artifact_retention", Scope::WorkflowsRead),
765 ("delete_artifact", Scope::WorkflowsWrite),
766 ("set_artifact_retention", Scope::WorkflowsWrite),
Merge checks: statuses and check runs on every commit767 // Checks: statuses, check runs and check suites on commits.
768 ("list_commit_statuses", Scope::ChecksRead),
769 ("get_combined_status", Scope::ChecksRead),
770 ("list_check_runs_for_ref", Scope::ChecksRead),
771 ("get_check_run", Scope::ChecksRead),
772 ("list_check_run_annotations", Scope::ChecksRead),
773 ("list_check_suites_for_ref", Scope::ChecksRead),
774 ("get_check_suite", Scope::ChecksRead),
775 ("create_commit_status", Scope::ChecksWrite),
776 ("create_check_run", Scope::ChecksWrite),
777 ("update_check_run", Scope::ChecksWrite),
778 ("rerequest_check_run", Scope::ChecksWrite),
779 ("rerequest_check_suite", Scope::ChecksWrite),
Merge branch 'main' into worktree-agent-a69aeabc4b0deeb97780 // Deployments, wherever they run: reading them, and reporting them.
781 ("list_deployments", Scope::DeploymentsRead),
782 ("get_deployment", Scope::DeploymentsRead),
783 ("list_deployment_statuses", Scope::DeploymentsRead),
784 ("list_environments", Scope::DeploymentsRead),
785 ("get_environment", Scope::DeploymentsRead),
786 ("create_deployment", Scope::DeploymentsWrite),
787 ("create_deployment_status", Scope::DeploymentsWrite),
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step788 // Memory and the context hub.
789 ("recall", Scope::MemoryRead),
790 ("search_context", Scope::MemoryRead),
791 ("get_entity", Scope::MemoryRead),
792 ("get_context", Scope::MemoryRead),
793 ("remember", Scope::MemoryWrite),
794 // Who has access.
795 ("list_collaborators", Scope::AccessRead),
796 ("get_collaborator_permission", Scope::AccessRead),
797 ("list_repo_invitations", Scope::AccessRead),
798 ("list_outside_collaborators", Scope::AccessRead),
799 ("add_collaborator", Scope::AccessAdmin),
800 ("update_collaborator", Scope::AccessAdmin),
801 ("remove_collaborator", Scope::AccessAdmin),
802 ("revoke_repo_invitation", Scope::AccessAdmin),
803 ("set_base_permission", Scope::AccessAdmin),
Teams and CODEOWNERS, labels and milestones, dependency updates, the security suite, and a clearer top bar804 ("set_team_repo", Scope::AccessAdmin),
805 ("remove_team_repo", Scope::AccessAdmin),
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step806 // Webhooks.
807 ("list_webhooks", Scope::WebhooksRead),
808 ("list_webhook_deliveries", Scope::WebhooksRead),
809 ("create_webhook", Scope::WebhooksAdmin),
810 ("update_webhook", Scope::WebhooksAdmin),
811 ("delete_webhook", Scope::WebhooksAdmin),
812 ("ping_webhook", Scope::WebhooksAdmin),
813 ("redeliver_webhook", Scope::WebhooksAdmin),
814 // Secrets and variables.
815 ("list_actions_secrets", Scope::SecretsRead),
816 ("list_actions_variables", Scope::SecretsRead),
817 ("set_actions_secret", Scope::SecretsAdmin),
818 ("delete_actions_secret", Scope::SecretsAdmin),
819 ("set_actions_variable", Scope::SecretsAdmin),
820 ("delete_actions_variable", Scope::SecretsAdmin),
Fast pages, required checks on the branch, self-hosted runners, honest incidents821 // Self-hosted runners.
822 ("list_runners", Scope::RunnersRead),
823 ("list_runner_groups", Scope::RunnersRead),
824 ("get_runner_settings", Scope::RunnersRead),
825 ("create_runner_registration_token", Scope::RunnersAdmin),
826 ("remove_runner", Scope::RunnersAdmin),
827 ("create_runner_group", Scope::RunnersAdmin),
828 ("update_runner_group", Scope::RunnersAdmin),
829 ("delete_runner_group", Scope::RunnersAdmin),
830 ("update_runner_settings", Scope::RunnersAdmin),
Merge the AI Gateway: Anthropic's Messages API on a workspace's tokens831 // The AI Gateway. Sending a request to a model needs `models:write`,
832 // checked by the model proxy at models.g1t.sh, not here.
833 ("list_gateway_requests", Scope::ModelsRead),
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step834];
835
836/// Operations any token may use: saying who it is.
837pub const NO_SCOPE: &[&str] = &["whoami"];
838
839/// The scope `operation` needs. `None` for one in [`NO_SCOPE`]; an
840/// operation in neither list needs full access.
841pub fn scope_for(operation: &str) -> Option<Scope> {
842 OPERATIONS
843 .iter()
844 .find(|(name, _)| *name == operation)
845 .map(|(_, scope)| *scope)
846}
847
848/// What a token needs for `operation` with this input beyond its own
849/// scope: starting agents from an operation that can, and making a
850/// repository public or private.
851pub fn extra_scopes(operation: &str, input: &serde_json::Value) -> Vec<Scope> {
852 let mut extra = Vec::new();
853 let assigns = input["assign"].as_bool() == Some(true)
854 || input["agent"].as_bool() == Some(true)
855 || input["assign_agent"].as_bool() == Some(true);
856 if assigns && matches!(operation, "apply_plan" | "import_issue" | "create_issue") {
857 extra.push(Scope::AgentsRun);
858 }
Teams and CODEOWNERS, labels and milestones, dependency updates, the security suite, and a clearer top bar859 // Fixing an alert opens an issue and puts g1t on it.
860 if operation == "fix_security_alert" {
861 extra.extend([Scope::IssuesWrite, Scope::AgentsRun]);
862 }
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step863 // Opening the issue an agent is put on.
864 if operation == "delegate" {
865 extra.push(Scope::IssuesWrite);
866 }
Git storage hardened, pages in tens of milliseconds, honest security alerts, and costs reconciled daily867 // A workspace's base permission is who has access.
868 if operation == "update_workspace" && input.get("base_permission").is_some_and(|v| !v.is_null()) {
869 extra.push(Scope::AccessAdmin);
870 }
Merge checks: statuses and check runs on every commit871 // Asking a g1t Actions job or run to run again reruns its workflow.
872 if matches!(operation, "rerequest_check_run" | "rerequest_check_suite")
873 && input["id"].as_str().is_some_and(|id| id.starts_with("job_") || id.starts_with("run_"))
874 {
875 extra.push(Scope::WorkflowsWrite);
876 }
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step877 if operation == "update_repo" && (input.get("private").is_some_and(|v| !v.is_null()) || input.get("default_branch").is_some_and(|v| !v.is_null())) {
878 extra.push(Scope::RepoAdmin);
879 }
880 extra
881}
882
883/// The scopes a call needs, its own first.
884pub fn needed(operation: &str, input: &serde_json::Value) -> Vec<Scope> {
885 scope_for(operation)
886 .into_iter()
887 .chain(extra_scopes(operation, input))
888 .collect()
889}
890
891/// Whether `access` may use `operation` with `input`. The person's (or
892/// workspace's) role is checked after this, by the service that owns what
893/// was asked about.
894pub fn decide(access: &TokenAccess, operation: &str, input: &serde_json::Value) -> Decision {
895 let rule = if access.legacy { "token:legacy" } else { "token:scope" };
896 if access.scopes.is_some() {
897 let known = NO_SCOPE.contains(&operation) || scope_for(operation).is_some();
898 if !known {
899 return Decision::deny("token:scope", format!("This access token cannot use {operation}: it needs full access."));
900 }
901 if let Some(missing) = needed(operation, input).into_iter().find(|scope| !access.allows(*scope)) {
902 return Decision::deny(
903 "token:scope",
904 format!("This access token needs the {} scope to use {operation}.", missing.as_str()),
905 );
906 }
907 }
908 Decision::allow(rule)
909}
910
911/// Whether a token may clone or fetch (`write` false), or push to (`write`
912/// true), a repository with git. `public` is whether anyone may read it,
913/// which needs no scope.
914pub fn decide_git(access: &TokenAccess, write: bool, public: bool) -> Decision {
915 let needed = if write { Scope::CodeWrite } else { Scope::CodeRead };
916 if !access.allows(needed) && (write || !public) {
917 return Decision::deny(
918 "token:scope",
919 format!("This access token needs the {} scope to {} with git.", needed.as_str(), if write { "push" } else { "clone or fetch a private repository" }),
920 );
921 }
922 Decision::allow(if access.legacy { "token:legacy" } else { "token:scope" })
923}
924
Packages, with a container registry on g1t.sh; workspaces deleted whole and kept 30 days; Members for every member925/// Whether a token may pull (`Level::Read`), push or publish
926/// (`Level::Write`), or delete (`Level::Delete`) packages. `public` is
927/// whether anyone may pull the package, which needs no scope.
928pub fn decide_packages(access: &TokenAccess, level: Level, public: bool) -> Decision {
929 let (needed, doing) = match level {
930 Level::Read => (Scope::PackagesRead, "pull a private package"),
931 Level::Delete | Level::Admin => (Scope::PackagesDelete, "delete packages"),
932 Level::Write | Level::Run => (Scope::PackagesWrite, "push or publish packages"),
933 };
934 if !access.allows(needed) && !(level == Level::Read && public) {
935 return Decision::deny(
936 "token:scope",
937 format!("This access token needs the {} scope to {doing}.", needed.as_str()),
938 );
939 }
940 Decision::allow(if access.legacy { "token:legacy" } else { "token:scope" })
941}
942
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step943#[cfg(test)]
944mod tests {
945 use super::*;
946 use serde_json::json;
947
948 fn token(scopes: &[Scope]) -> TokenAccess {
949 TokenAccess {
950 token_id: "tok_1".to_owned(),
951 scopes: Some(scopes.iter().map(|scope| scope.as_str().to_owned()).collect()),
952 legacy: false,
Merge the AI Gateway: Anthropic's Messages API on a workspace's tokens953 name: None,
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step954 }
955 }
956
957 #[test]
958 fn every_scope_reads_back_and_belongs_to_a_resource() {
959 for scope in Scope::ALL {
960 assert_eq!(Scope::parse(scope.as_str()), Some(scope));
961 assert!(scope.as_str().starts_with(scope.resource().as_str()));
962 assert!(scope.includes(scope));
963 }
964 assert_eq!(Scope::parse(" Issues:Write "), Some(Scope::IssuesWrite));
965 assert_eq!(Scope::parse("issues"), None);
966 }
967
968 #[test]
969 fn a_higher_level_includes_the_lower_ones_of_its_resource_only() {
970 assert!(Scope::RepoAdmin.includes(Scope::RepoRead));
971 assert!(Scope::RepoAdmin.includes(Scope::RepoWrite));
972 assert!(Scope::IssuesWrite.includes(Scope::IssuesRead));
973 assert!(!Scope::IssuesRead.includes(Scope::IssuesWrite));
974 assert!(!Scope::RepoAdmin.includes(Scope::CodeWrite));
975 assert!(!Scope::PullRequestsWrite.includes(Scope::IssuesWrite));
976 }
977
978 #[test]
979 fn operations_are_listed_once_and_never_also_free() {
980 let mut seen = std::collections::HashSet::new();
981 for (name, _) in OPERATIONS {
982 assert!(seen.insert(*name), "{name} twice");
983 assert!(!NO_SCOPE.contains(name), "{name}");
984 }
985 }
986
987 #[test]
988 fn scopes_are_parsed_from_oauth_text_leaving_out_unknown_ones() {
989 assert_eq!(
990 parse_scopes("issues:write repo:read,bogus:thing issues:write"),
991 vec![Scope::RepoRead, Scope::IssuesWrite]
992 );
993 assert_eq!(scopes_text(&[Scope::RepoRead, Scope::IssuesWrite]), "repo:read issues:write");
994 }
995
996 #[test]
997 fn the_oauth_default_is_the_agent_preset_and_never_admin() {
998 let scopes = oauth_default();
999 assert!(scopes.contains(&Scope::IssuesWrite));
1000 assert!(scopes.contains(&Scope::PullRequestsWrite));
1001 assert!(scopes.contains(&Scope::AgentsRun));
1002 assert!(scopes.iter().all(|scope| !scope.dangerous()), "{scopes:?}");
1003 for read in Scope::ALL.into_iter().filter(|scope| scope.level() == Level::Read) {
Fast pages, required checks on the branch, self-hosted runners, honest incidents1004 // Every read but the machines work runs on.
1005 assert_eq!(scopes.contains(&read), read != Scope::RunnersRead, "{read:?}");
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step1006 }
1007 assert!(Preset::ReadOnly.scopes().unwrap().iter().all(|scope| scope.level() == Level::Read));
1008 assert_eq!(Preset::Full.scopes(), None);
1009 }
1010
1011 #[test]
Usage, Billing settings and prepaid AI credit; fixes from the UX audit1012 fn billing_is_read_by_presets_and_changed_by_none_but_full_access() {
1013 assert!(Preset::ReadOnly.scopes().unwrap().contains(&Scope::BillingRead));
1014 for preset in [Preset::ReadOnly, Preset::Agent, Preset::Ci] {
1015 assert!(!preset.scopes().unwrap().contains(&Scope::BillingWrite), "{}", preset.as_str());
1016 }
1017 assert_eq!(scope_for("set_budget"), Some(Scope::BillingWrite));
1018 assert_eq!(scope_for("buy_ai_credit"), Some(Scope::BillingWrite));
1019 assert_eq!(scope_for("get_usage"), Some(Scope::BillingRead));
1020 let reader = token(&[Scope::BillingRead]);
1021 assert!(decide(&reader, "list_invoices", &json!({})).allowed);
1022 assert!(decide(&reader, "set_budget", &json!({})).reason.unwrap().contains("billing:write"));
1023 }
1024
1025 #[test]
Merge the AI Gateway: Anthropic's Messages API on a workspace's tokens1026 fn the_ai_gateway_spends_only_with_models_write_which_no_preset_gives() {
1027 // Reading the log is a read like any other.
1028 assert_eq!(scope_for("list_gateway_requests"), Some(Scope::ModelsRead));
1029 assert!(Preset::ReadOnly.scopes().unwrap().contains(&Scope::ModelsRead));
1030 // Sending requests spends the workspace's AI credit: chosen on purpose.
1031 for preset in [Preset::ReadOnly, Preset::Agent, Preset::Ci] {
1032 assert!(!preset.scopes().unwrap().contains(&Scope::ModelsWrite), "{}", preset.as_str());
1033 }
1034 assert!(Scope::ModelsWrite.includes(Scope::ModelsRead));
1035 assert!(!Scope::ModelsWrite.dangerous());
1036 assert!(token(&[Scope::ModelsWrite]).allows(Scope::ModelsWrite));
1037 assert!(!token(&[Scope::BillingWrite]).allows(Scope::ModelsWrite));
1038 assert!(TokenAccess::full().allows(Scope::ModelsWrite));
1039 }
1040
1041 #[test]
Merge checks: statuses and check runs on every commit1042 fn checks_are_reported_with_checks_write_which_ci_gets() {
1043 assert_eq!(scope_for("create_check_run"), Some(Scope::ChecksWrite));
1044 assert_eq!(scope_for("create_commit_status"), Some(Scope::ChecksWrite));
1045 assert_eq!(scope_for("list_check_runs_for_ref"), Some(Scope::ChecksRead));
1046 let ci = Preset::Ci.scopes().unwrap();
1047 assert!(ci.contains(&Scope::ChecksWrite));
1048 assert!(!Preset::Agent.scopes().unwrap().contains(&Scope::ChecksWrite));
1049 let reporter = token(&[Scope::ChecksWrite]);
1050 assert!(decide(&reporter, "update_check_run", &json!({ "id": "cr_1" })).allowed);
1051 assert!(decide(&reporter, "rerequest_check_run", &json!({ "id": "cr_1" })).allowed);
1052 // A g1t Actions job runs again as its workflow does.
1053 let refused = decide(&reporter, "rerequest_check_run", &json!({ "id": "job_1" }));
1054 assert!(refused.reason.unwrap().contains("workflows:write"));
1055 }
1056
1057 #[test]
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step1058 fn a_legacy_token_can_do_everything() {
1059 let legacy = TokenAccess { legacy: true, ..TokenAccess::full() };
1060 for (operation, _) in OPERATIONS {
1061 assert!(decide(&legacy, operation, &json!({})).allowed, "{operation}");
1062 }
1063 assert_eq!(decide(&legacy, "delete_repo", &json!({})).rule, "token:legacy");
1064 }
1065
1066 #[test]
1067 fn a_missing_scope_is_named() {
1068 let read = token(&[Scope::IssuesRead]);
1069 assert!(decide(&read, "get_issue", &json!({})).allowed);
1070 assert!(decide(&read, "whoami", &json!({})).allowed);
1071 let refused = decide(&read, "create_issue", &json!({}));
1072 assert!(!refused.allowed);
1073 assert_eq!(refused.reason.as_deref(), Some("This access token needs the issues:write scope to use create_issue."));
1074 // An operation the table does not know needs full access.
1075 assert!(!decide(&read, "something_new", &json!({})).allowed);
1076 }
1077
1078 #[test]
1079 fn starting_agents_from_another_operation_needs_agents_run() {
1080 let writer = token(&[Scope::IssuesWrite]);
1081 assert!(decide(&writer, "apply_plan", &json!({})).allowed);
1082 let refused = decide(&writer, "apply_plan", &json!({ "assign": true }));
1083 assert!(refused.reason.unwrap().contains("agents:run"));
1084 let maintainer = token(&[Scope::RepoWrite]);
1085 assert!(decide(&maintainer, "update_repo", &json!({ "description": "x" })).allowed);
1086 assert!(!decide(&maintainer, "update_repo", &json!({ "private": true })).allowed);
1087 }
1088
1089 #[test]
Git storage hardened, pages in tens of milliseconds, honest security alerts, and costs reconciled daily1090 fn a_workspaces_base_permission_needs_access_admin_too() {
1091 let admin = token(&[Scope::WorkspaceAdmin]);
1092 assert!(decide(&admin, "update_workspace", &json!({ "name": "Acme" })).allowed);
1093 let refused = decide(&admin, "update_workspace", &json!({ "name": "Acme", "base_permission": "read" }));
1094 assert!(refused.reason.unwrap().contains("access:admin"));
1095 let both = token(&[Scope::WorkspaceAdmin, Scope::AccessAdmin]);
1096 assert!(decide(&both, "update_workspace", &json!({ "base_permission": "read" })).allowed);
1097 assert!(!decide(&token(&[Scope::WorkspaceRead]), "update_workspace", &json!({ "name": "Acme" })).allowed);
1098 }
1099
1100 #[test]
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step1101 fn delegating_needs_both_agents_and_issues() {
1102 let agents = token(&[Scope::AgentsRun]);
1103 assert!(decide(&agents, "delegate", &json!({})).reason.unwrap().contains("issues:write"));
1104 let both = token(&[Scope::AgentsRun, Scope::IssuesWrite]);
1105 assert!(decide(&both, "delegate", &json!({})).allowed);
1106 }
1107
1108 #[test]
1109 fn git_push_needs_code_write_and_private_reads_need_code_read() {
1110 let reader = token(&[Scope::CodeRead]);
1111 assert!(decide_git(&reader, false, false).allowed);
1112 let refused = decide_git(&reader, true, false);
1113 assert!(!refused.allowed);
1114 assert!(refused.reason.unwrap().contains("code:write"));
1115 let issues = token(&[Scope::IssuesWrite]);
1116 assert!(!decide_git(&issues, false, false).allowed);
1117 assert!(decide_git(&issues, false, true).allowed, "public code needs no scope");
1118 assert!(!decide_git(&issues, true, true).allowed, "pushing to public code still needs code:write");
1119 let writer = token(&[Scope::CodeWrite]);
1120 assert!(decide_git(&writer, true, false).allowed);
1121 assert!(decide_git(&writer, false, false).allowed, "code:write includes code:read");
1122 assert!(decide_git(&TokenAccess::full(), true, false).allowed);
1123 }
1124
1125 #[test]
Packages, with a container registry on g1t.sh; workspaces deleted whole and kept 30 days; Members for every member1126 fn packages_need_their_own_scopes_and_public_pulls_none() {
1127 let reader = token(&[Scope::PackagesRead]);
1128 assert!(decide_packages(&reader, Level::Read, false).allowed);
1129 assert!(!decide_packages(&reader, Level::Write, false).allowed);
1130 let code = token(&[Scope::CodeWrite]);
1131 assert!(!decide_packages(&code, Level::Read, false).allowed, "code scopes are not package scopes");
1132 assert!(decide_packages(&code, Level::Read, true).allowed, "public packages pull with any token");
1133 let writer = token(&[Scope::PackagesWrite]);
1134 assert!(decide_packages(&writer, Level::Write, false).allowed);
1135 assert!(decide_packages(&writer, Level::Read, false).allowed, "packages:write includes packages:read");
1136 let refused = decide_packages(&writer, Level::Delete, false);
1137 assert!(refused.reason.unwrap().contains("packages:delete"));
1138 assert!(decide_packages(&token(&[Scope::PackagesDelete]), Level::Write, false).allowed);
1139 assert!(Scope::PackagesDelete.dangerous());
1140 // Tokens made before these scopes, and full-access ones, keep working.
1141 let legacy = TokenAccess { legacy: true, ..TokenAccess::full() };
1142 assert!(decide_packages(&legacy, Level::Delete, false).allowed);
1143 assert!(decide_packages(&TokenAccess::full(), Level::Write, false).allowed);
1144 }
1145
1146 #[test]
Thirteen MCP tools and classic token scopes; agents rate their confidence and can be put on an issue in one step1147 fn token_access_travels_as_json() {
1148 let access = token(&[Scope::IssuesRead]);
1149 let wire = serde_json::to_value(&access).unwrap();
1150 assert_eq!(wire["scopes"], json!(["issues:read"]));
1151 assert!(wire.get("resources").is_none());
1152 let back: TokenAccess = serde_json::from_value(wire).unwrap();
1153 assert_eq!(back, access);
1154 let full: TokenAccess = serde_json::from_value(json!({})).unwrap();
1155 assert!(full.is_full());
1156 // A reach written by an older version is ignored: a token reaches
1157 // whatever its owner can.
1158 let older: TokenAccess = serde_json::from_value(json!({
1159 "token_id": "tok_1",
1160 "scopes": ["issues:read"],
1161 "resources": { "kind": "repositories", "repositories": ["acme/rocket"] },
1162 }))
1163 .unwrap();
1164 assert_eq!(older, access);
1165 }
1166
1167 /// The site's copy of the table, `packages/contracts/src/scopes.ts`,
1168 /// lists the same scopes in the same order, the same operations with
1169 /// the same scopes, and the same presets.
1170 #[test]
1171 fn the_typescript_mirror_has_the_same_table() {
1172 let ts = include_str!("../../../packages/contracts/src/scopes.ts");
1173 let section = |start: &str| {
1174 ts.split_once(start)
1175 .and_then(|(_, rest)| rest.split_once("] as const"))
1176 .map(|(table, _)| table)
1177 .unwrap_or_else(|| panic!("{start} in scopes.ts"))
1178 };
1179 let scopes: Vec<&str> = section("export const SCOPES = [")
1180 .lines()
1181 .filter_map(|line| line.split_once("scope: \"").and_then(|(_, rest)| rest.split_once('"')).map(|(scope, _)| scope))
1182 .collect();
1183 let expected: Vec<&str> = Scope::ALL.iter().map(|scope| scope.as_str()).collect();
1184 assert_eq!(scopes, expected);
1185 let operations: Vec<(String, String)> = section("export const OPERATION_SCOPES = [")
1186 .lines()
1187 .filter_map(|line| {
1188 let mut quoted = line.split('"').skip(1).step_by(2);
1189 Some((quoted.next()?.to_owned(), quoted.next()?.to_owned()))
1190 })
1191 .collect();
1192 let expected: Vec<(String, String)> = OPERATIONS
1193 .iter()
1194 .map(|(name, scope)| ((*name).to_owned(), scope.as_str().to_owned()))
1195 .collect();
1196 assert_eq!(operations, expected);
1197 for preset in Preset::ALL {
1198 let list = section(&format!("{}: [", preset.as_str()));
1199 let mirrored: Vec<&str> = list
1200 .split(',')
1201 .map(|item| item.trim().trim_matches('"'))
1202 .filter(|item| !item.is_empty())
1203 .collect();
1204 let expected: Vec<&str> = preset
1205 .scopes()
1206 .map(|scopes| scopes.iter().map(|scope| scope.as_str()).collect())
1207 .unwrap_or_else(|| vec!["*"]);
1208 assert_eq!(mirrored, expected, "{}", preset.as_str());
1209 }
1210 }
1211}

This file's history is long; its oldest lines are credited to the oldest commit read.