crates/contracts/src · flagon-io/g1t

workAgents in any harness coordinate through one inbox, the merge queue lands only what an attempt itself tested, and workspaces have Insights. Each session keeps the coordination updates that concern it (session_updates, migration 0038): claims, releases, hand-offs and expiries, overlaps, dependency changes, decisions, integration results and sessions starting and ending, about the issues and pull requests it is on or tied to, or every one while it is on nothing yet; never its own start, end or claims. An update is written in the same batch as the change it announces, guarded so it exists only once the change is made, with an id derived from the change, so a retry or two requests at once deliver it once; a blocker clearing and queue results follow their change under a key that names it. get_session's `updates` lists up to 100, oldest first, and reading never clears them; record_session_entries' `acknowledge` clears exact IDs for that session only and says how many were its (`acknowledged`). The CLI has `g1t session updates` and `g1t session acknowledge ID --note` (the decision note and the acknowledgement in one idempotent call, `--session` over G1T_SESSION), and those and `heartbeat` say which files the working copy changed (git status; G1T_TOUCH=0 turns it off), so overlapping edits are noticed whichever harness makes them; a pair of sessions is told again when another file in common joins. The Claude Code hooks fetch the inbox in the background, show it at every prompt and before tools when it changes or every two minutes, refresh after an acknowledgement, and never acknowledge. The AGENTS.md snippet, the skills, MCP's instructions and llms.txt say the same steps for every harness: scope claims (Scope, Interfaces, Depends on, Checks), proceed on complementary work, agree ownership before overlapping edits, record public summaries and evidence, never private reasoning. Claims resolve ownership in one statement, so simultaneous claims by one session converge. The merge queue: every attempt has its own id and branch (0037); a claimed build, an accepted build, a report and a callback each check their attempt; a dispatch registers its runs first and Actions fails the dispatch when it cannot; an attempt is decided only by its own runs (and other tools' statuses set after it was claimed), never an earlier attempt's or the pull request's own on the same commit (0040 keeps what it built and how many runs it registered); landing re-reads the rules, honours a bypass asked for when queued (0039), checks no predecessor remains, carries the exact tested commit to Git's compare-and-swap, records publication intent that fences removal, closing, drafting and changing base, resumes after a crash, and a publisher that lost a race no longer undoes the winner's landing; a retried attempt's branch is deleted; an entry whose branch or pull request is gone leaves the queue with a note instead of blocking those behind it; diagnosis rebuilds exactly what was tested and names a failing combination, not a proven cause. A session's page shows what awaits its agent as the coordination feed does. Workspace Insights (/:owner/-/insights): delivery metrics across a workspace's readable repositories (work service delivery_insights). Deploy Work migrations 0037 to 0040 first (the pipeline does), then Repos and Actions before Work, and the site with Work. Checked: cargo test for g1t-work (143), g1t-repos (311), g1t-actions-service (108), g1t-api (201), g1t (149), g1t-rules (72), g1t-contracts (306) and wasm32 builds; cargo check --workspace; the conformance suite against freshly built API and Work Workers (5, none skipped); the Claude Code hooks (49); the site's typecheck, tests (919) and a production build; the deploy tool's 92 tests; the docs build; and two unscripted local rehearsals in which two concurrent agents, given only a task and AGENTS.md, found each other, agreed a split and made complementary edits.2d ago

about.rsMerge remote-tracking branch 'origin/claude/cool-newton-7nm8r9' into claude/cool-newton-7nm8r95d ago

access.rsOne kind of access token; presence and status; usernames keep their case; the tour is a miniature of the real app; icons for password managers1w ago

account_deletion.rsOne kind of access token; presence and status; usernames keep their case; the tour is a miniature of the real app; icons for password managers1w ago

accounts.rsMerge main into rm-agents: mirroring removal, open signups, free-tier limits, the g1t CLI and runner fixes, with the agent removal kept. The CLI's table of operations is regenerated from the API without the agent, customize and AI credit operations; no CLI command used them.4d ago

actions.rsAsk names people by their g1t username, knows what shipped means, and picks its model by itself. Asked "What shipped today?", Ask made "@chasepierce" out of a commit's git author name, corrected itself mid-answer, and said nothing had reached production though the repository's own Deploy run had: commit listings carried only the git name, the prompt never said which field is a username, and activity counted only g1t.page deployments as production. Commits from list_commits now say `author: { name, username }`, the username from identity's address owners (a confirmed or noreply address; g1t's own are g1t), never the address, and activity's landed commits carry `username` too; the prompt mentions only usernames tools returned, never turns a git name or address into one, and answers once, without "wait" or corrections. Shipping, to Ask, is reaching the default branch and then production: a successful run of a repository's deploy workflow on its default branch (a workflow whose name or file says deploy, not preview) now counts as a production deployment beside g1t.page's, each landed commit says whether it is live (and by which deploy), deploying, failed or not deployed (a deploy carries everything before it on the branch), summary.shipped counts them, and list_runs takes since and until; releases and tags are one more signal, and today is the start of your day in your time zone. Ask's model is Auto by default (migration 0010 turns it on for every Ask route nobody chose a model for: the old picker started on the provider's first model, Haiku for Anthropic). Auto sends each question to a model of the workspace's provider without asking another model: a short lookup to a fast one, a question about why, how, what shipped, changes, failures, a time range, or a long conversation to a standard one; tiers come from the models' names, so it works for every provider, the largest only when nothing else is there, and an Azure deployment or an endpoint without a list keeps its own model. The answer says which ("Anthropic · Auto · claude-sonnet-5-5"); an explicit pick still wins. The Ask guide has Auto and what Ask means by shipped. Checked on this change alone (a worktree of the commit before it): cargo test for g1t-integrations (70, new: Auto's routing for each provider, prompt rules), g1t-api (204, new: commit authors with and without an account, shipped states), g1t-actions-service (108), g1t-contracts (306) and the CLI (149); cargo check --workspace; wasm32 builds of api, integrations and actions; the site's typecheck (tsc -b --force), its Ask tests and a production build; chat's tests and typecheck; the deploy tool's tests.2d ago

advisories.rsMerge remote-tracking branch 'origin/claude/cool-newton-7nm8r9' into claude/cool-newton-7nm8r95d ago

agent_sessions.rsg1t coordinates the work and the person's own sessions do it: Ask g1t reads g1t with the same tools any agent gets, answers in one conversation shared with the @g1t direct message in Chat, and hands work to sessions in the person's own harness, whose subagents g1t records, ties to issues and pull requests, and lets see each other.4d ago

agents.rsMerge branch 'worktree-agent-a4babc5b5815f6f3a'3d ago

audit.rsMerge main into Artifacts Phase 21w ago

backups.rsThe docs folder is gone, and what it held lives where people read it: how a self-hosted g1t runs and how to deploy g1t to Cloudflare are pages on docs.g1t.sh under Run g1t yourself, and speed, rate limits and operating g1t.sh are sections of CONTRIBUTING.md; code that cited a file in docs/ now points to the page or section that covers it, or says what it means itself, and applied migrations and the runner images are left as they were.1w ago

billing.rsBilling is ready to take real money, and stays on Stripe's test key until the switch. The g1t plan is $21 a month, all-in, with $10 of usage included: the card processing fee is part of the price, and no payment carries a fee line any more (old fee rows still read as they were). Card, billing address, tax IDs, invoices and cancelling are on Stripe's billing page (Manage billing), whose configuration billing keeps in step; Checkout asks for the billing address and a tax ID; g1t's Billing page keeps usage, the spend limit, budget alerts and prepaying. Members other than owners and billing managers no longer see billing details or invoices, from the page or the API.3d ago

checks.rsA required check named for a workflow job, as GitHub's branch rules name them, is met by that job. Each g1t Actions job on a pull request's head counts as a check of its name (`test`, `build (ubuntu-latest, 22)`, or `deploy / build` for a called workflow's job), from the latest attempt of its workflow: skipped passes, cancelled fails, and a job still going holds the merge. So a ruleset or required check copied from GitHub needs no renaming, and one that names the workflow, such as `CI`, is met by the whole run as before. This holds in the merge box, for the merge queue and for agents' merges; jobs count only toward required checks, so a `continue-on-error` job that fails doesn't send an agent back. The pull requests guide says so.5d ago

codeowners.rsTeams and CODEOWNERS, labels and milestones, dependency updates, the security suite, and a clearer top bar1w ago

coding_agents.rsMerge remote-tracking branch 'origin/claude/cool-newton-7nm8r9' into claude/cool-newton-7nm8r95d ago

credentials.rsMerge branch 'worktree-agent-a4babc5b5815f6f3a'3d ago

datasets.rsThe docs folder is gone, and what it held lives where people read it: how a self-hosted g1t runs and how to deploy g1t to Cloudflare are pages on docs.g1t.sh under Run g1t yourself, and speed, rate limits and operating g1t.sh are sections of CONTRIBUTING.md; code that cited a file in docs/ now points to the page or section that covers it, or says what it means itself, and applied migrations and the runner images are left as they were.1w ago

decisions.rs`why` no longer hands out commit authors' email addresses. Why a line is the way it is (API why, MCP repository why) returned the blamed commit as git stores it, author address included, to anyone who can read the repository. Its commit is now a WhyCommit: the author is named as git wrote them and by `username`, the g1t account their address belongs to (a confirmed or noreply address, or g1t's own), as list_commits already does, and the address is never returned. The guide, the operation's description, its example and the OpenAPI document say so. Checked: cargo test for g1t-contracts (new: an author is named by account, never by address) and g1t-api (204, the OpenAPI document regenerated), the CLI's (149), cargo check --workspace and g1t-api's wasm32 build.2d ago

deploy_keys.rsMerge branch 'worktree-agent-a5a5de74d8863d554' into worktree-agent-a16631325aecf58ca1w ago

events.rsMerge branch 'worktree-agent-a4babc5b5815f6f3a'3d ago

folios.rsThe contracts no longer describe g1t's own agents, and workspaces can keep their AI providers. Every consumer had already stopped using them, so the TypeScript and Rust contracts drop the workspace-agent, ability, look, skill-library, context, runner-agent and mention modules, the chat and docs RPCs agents posted and read through, the folio agent methods, the agent_waiting notification, agent.asked and memory.changed events, AI credit, token and gateway billing types, model routes and sessions, team agents, agent tokens, plan, review, message and confidence types, command rules, cost caps and confidence rules. The agents:run and models:* scopes are no longer offered by any preset, OAuth default or permission form, but a stored token or grant that names them still reads; and the removed operations are gone from the scope table, so the API's check that every operation has exactly one scope is strict again. What stored data needs in order to parse (old run and credential kinds, an agent team lead, retired rules and webhook events, old guardrail and integration settings) still parses, and a run's token is still PrincipalKind::Agent, which now means a run's token and the @g1t rows written before. The customize, marketplace, skills and claude-plugin contracts stay only until the mobile app drops its Customize screens.4d ago

github.rsMerge main into rm-agents: mirroring removal, open signups, free-tier limits, the g1t CLI and runner fixes, with the agent removal kept. The CLI's table of operations is regenerated from the API without the agent, customize and AI credit operations; no CLI command used them.4d ago

guardrails.rsMerge branch 'worktree-agent-ac950b9c891220ad3'3d ago

identity.rsMerge branch 'worktree-agent-a6da31375506fc440'3d ago

ids.rsHome says what people did as well as what agents did. Since you were last here now has a People column and an Agents column: who pushed how many commits to which projects, pull requests opened, merged and reviewed, issues opened and closed, docs edited, messages sent and deploys that went out, each line a link to where those are listed, with the agents' acceptance (first time, after review, didn't finish) kept as a row of their column; the sentence under the heading sums it up honestly, and says when a part of g1t could not be read, or that the span was quiet. Landed counts merged pull requests, commits pushed straight to a default branch, production deploys that went live, releases and packages, newest first; Running now adds workflow runs. Behind it, the events service answers an activity digest over a span in one round trip from its existing indexes, every push now records how many commits it carried, and chat counts the messages sent in the conversations you can read. The Home guide defines every line, and says how this scales.6d ago

inbox.rsMerge remote-tracking branch 'origin/claude/cool-newton-7nm8r9' into claude/cool-newton-7nm8r95d ago

integrations.rsAsk names people by their g1t username, knows what shipped means, and picks its model by itself. Asked "What shipped today?", Ask made "@chasepierce" out of a commit's git author name, corrected itself mid-answer, and said nothing had reached production though the repository's own Deploy run had: commit listings carried only the git name, the prompt never said which field is a username, and activity counted only g1t.page deployments as production. Commits from list_commits now say `author: { name, username }`, the username from identity's address owners (a confirmed or noreply address; g1t's own are g1t), never the address, and activity's landed commits carry `username` too; the prompt mentions only usernames tools returned, never turns a git name or address into one, and answers once, without "wait" or corrections. Shipping, to Ask, is reaching the default branch and then production: a successful run of a repository's deploy workflow on its default branch (a workflow whose name or file says deploy, not preview) now counts as a production deployment beside g1t.page's, each landed commit says whether it is live (and by which deploy), deploying, failed or not deployed (a deploy carries everything before it on the branch), summary.shipped counts them, and list_runs takes since and until; releases and tags are one more signal, and today is the start of your day in your time zone. Ask's model is Auto by default (migration 0010 turns it on for every Ask route nobody chose a model for: the old picker started on the provider's first model, Haiku for Anthropic). Auto sends each question to a model of the workspace's provider without asking another model: a short lookup to a fast one, a question about why, how, what shipped, changes, failures, a time range, or a long conversation to a standard one; tiers come from the models' names, so it works for every provider, the largest only when nothing else is there, and an Azure deployment or an endpoint without a list keeps its own model. The answer says which ("Anthropic · Auto · claude-sonnet-5-5"); an explicit pick still wins. The Ask guide has Auto and what Ask means by shipped. Checked on this change alone (a worktree of the commit before it): cargo test for g1t-integrations (70, new: Auto's routing for each provider, prompt rules), g1t-api (204, new: commit authors with and without an account, shipped states), g1t-actions-service (108), g1t-contracts (306) and the CLI (149); cargo check --workspace; wasm32 builds of api, integrations and actions; the site's typecheck (tsc -b --force), its Ask tests and a production build; chat's tests and typecheck; the deploy tool's tests.2d ago

lib.rsg1t coordinates the work and the person's own sessions do it: Ask g1t reads g1t with the same tools any agent gets, answers in one conversation shared with the @g1t direct message in Chat, and hands work to sessions in the person's own harness, whose subagents g1t records, ties to issues and pull requests, and lets see each other.4d ago

live.rsg1t coordinates the work and the person's own sessions do it: Ask g1t reads g1t with the same tools any agent gets, answers in one conversation shared with the @g1t direct message in Chat, and hands work to sessions in the person's own harness, whose subagents g1t records, ties to issues and pull requests, and lets see each other.4d ago

members.rsBilling is ready to take real money, and stays on Stripe's test key until the switch. The g1t plan is $21 a month, all-in, with $10 of usage included: the card processing fee is part of the price, and no payment carries a fee line any more (old fee rows still read as they were). Card, billing address, tax IDs, invoices and cancelling are on Stripe's billing page (Manage billing), whose configuration billing keeps in step; Checkout asks for the billing address and a tax ID; g1t's Billing page keeps usage, the spend limit, budget alerts and prepaying. Members other than owners and billing managers no longer see billing details or invoices, from the page or the API.3d ago

names.rsMerge branch 'worktree-agent-ac950b9c891220ad3'3d ago

outcome.rsMerge branch 'mirroring' into artifacts-mode1w ago

packages.rsThe docs folder is gone, and what it held lives where people read it: how a self-hosted g1t runs and how to deploy g1t to Cloudflare are pages on docs.g1t.sh under Run g1t yourself, and speed, rate limits and operating g1t.sh are sections of CONTRIBUTING.md; code that cited a file in docs/ now points to the page or section that covers it, or says what it means itself, and applied migrations and the runner images are left as they were.1w ago

people.rsThe contracts no longer describe g1t's own agents, and workspaces can keep their AI providers. Every consumer had already stopped using them, so the TypeScript and Rust contracts drop the workspace-agent, ability, look, skill-library, context, runner-agent and mention modules, the chat and docs RPCs agents posted and read through, the folio agent methods, the agent_waiting notification, agent.asked and memory.changed events, AI credit, token and gateway billing types, model routes and sessions, team agents, agent tokens, plan, review, message and confidence types, command rules, cost caps and confidence rules. The agents:run and models:* scopes are no longer offered by any preset, OAuth default or permission form, but a stored token or grant that names them still reads; and the removed operations are gone from the scope table, so the API's check that every operation has exactly one scope is strict again. What stored data needs in order to parse (old run and credential kinds, an agent team lead, retired rules and webhook events, old guardrail and integration settings) still parses, and a run's token is still PrincipalKind::Agent, which now means a run's token and the @g1t rows written before. The customize, marketplace, skills and claude-plugin contracts stay only until the mobile app drops its Customize screens.4d ago

projects.rsProjects: what a workspace builds and runs, first on every page1w ago

repos.rsAgents in any harness coordinate through one inbox, the merge queue lands only what an attempt itself tested, and workspaces have Insights. Each session keeps the coordination updates that concern it (session_updates, migration 0038): claims, releases, hand-offs and expiries, overlaps, dependency changes, decisions, integration results and sessions starting and ending, about the issues and pull requests it is on or tied to, or every one while it is on nothing yet; never its own start, end or claims. An update is written in the same batch as the change it announces, guarded so it exists only once the change is made, with an id derived from the change, so a retry or two requests at once deliver it once; a blocker clearing and queue results follow their change under a key that names it. get_session's `updates` lists up to 100, oldest first, and reading never clears them; record_session_entries' `acknowledge` clears exact IDs for that session only and says how many were its (`acknowledged`). The CLI has `g1t session updates` and `g1t session acknowledge ID --note` (the decision note and the acknowledgement in one idempotent call, `--session` over G1T_SESSION), and those and `heartbeat` say which files the working copy changed (git status; G1T_TOUCH=0 turns it off), so overlapping edits are noticed whichever harness makes them; a pair of sessions is told again when another file in common joins. The Claude Code hooks fetch the inbox in the background, show it at every prompt and before tools when it changes or every two minutes, refresh after an acknowledgement, and never acknowledge. The AGENTS.md snippet, the skills, MCP's instructions and llms.txt say the same steps for every harness: scope claims (Scope, Interfaces, Depends on, Checks), proceed on complementary work, agree ownership before overlapping edits, record public summaries and evidence, never private reasoning. Claims resolve ownership in one statement, so simultaneous claims by one session converge. The merge queue: every attempt has its own id and branch (0037); a claimed build, an accepted build, a report and a callback each check their attempt; a dispatch registers its runs first and Actions fails the dispatch when it cannot; an attempt is decided only by its own runs (and other tools' statuses set after it was claimed), never an earlier attempt's or the pull request's own on the same commit (0040 keeps what it built and how many runs it registered); landing re-reads the rules, honours a bypass asked for when queued (0039), checks no predecessor remains, carries the exact tested commit to Git's compare-and-swap, records publication intent that fences removal, closing, drafting and changing base, resumes after a crash, and a publisher that lost a race no longer undoes the winner's landing; a retried attempt's branch is deleted; an entry whose branch or pull request is gone leaves the queue with a note instead of blocking those behind it; diagnosis rebuilds exactly what was tested and names a failing combination, not a proven cause. A session's page shows what awaits its agent as the coordination feed does. Workspace Insights (/:owner/-/insights): delivery metrics across a workspace's readable repositories (work service delivery_insights). Deploy Work migrations 0037 to 0040 first (the pipeline does), then Repos and Actions before Work, and the site with Work. Checked: cargo test for g1t-work (143), g1t-repos (311), g1t-actions-service (108), g1t-api (201), g1t (149), g1t-rules (72), g1t-contracts (306) and wasm32 builds; cargo check --workspace; the conformance suite against freshly built API and Work Workers (5, none skipped); the Claude Code hooks (49); the site's typecheck, tests (919) and a production build; the deploy tool's 92 tests; the docs build; and two unscripted local rehearsals in which two concurrent agents, given only a task and AGENTS.md, found each other, agreed a split and made complementary edits.2d ago

rules.rsThe contracts no longer describe g1t's own agents, and workspaces can keep their AI providers. Every consumer had already stopped using them, so the TypeScript and Rust contracts drop the workspace-agent, ability, look, skill-library, context, runner-agent and mention modules, the chat and docs RPCs agents posted and read through, the folio agent methods, the agent_waiting notification, agent.asked and memory.changed events, AI credit, token and gateway billing types, model routes and sessions, team agents, agent tokens, plan, review, message and confidence types, command rules, cost caps and confidence rules. The agents:run and models:* scopes are no longer offered by any preset, OAuth default or permission form, but a stored token or grant that names them still reads; and the removed operations are gone from the scope table, so the API's check that every operation has exactly one scope is strict again. What stored data needs in order to parse (old run and credential kinds, an agent team lead, retired rules and webhook events, old guardrail and integration settings) still parses, and a run's token is still PrincipalKind::Agent, which now means a run's token and the @g1t rows written before. The customize, marketplace, skills and claude-plugin contracts stay only until the mobile app drops its Customize screens.4d ago

runners.rsMerge branch 'worktree-agent-ac950b9c891220ad3'3d ago

scopes.rsMerge branch 'worktree-agent-a4babc5b5815f6f3a'3d ago

search.rsg1t is the stored author of what it opens; the person who asked is requested_by and keeps the author's rights1w ago

security.rsSecurity and version updates open no issues on their own: an update whose version cannot be raised fails with why on its alert and is tried again a day later, and one whose required checks fail stays open for a person to decide on; an agent takes one only when a person asks with Fix with g1t. That issue now names the lockfiles the package is locked in and the commands that show it is done, which were left empty ("Locked in: .") because the lockfiles were never passed in.4d ago

security_suite.rsThe contracts no longer describe g1t's own agents, and workspaces can keep their AI providers. Every consumer had already stopped using them, so the TypeScript and Rust contracts drop the workspace-agent, ability, look, skill-library, context, runner-agent and mention modules, the chat and docs RPCs agents posted and read through, the folio agent methods, the agent_waiting notification, agent.asked and memory.changed events, AI credit, token and gateway billing types, model routes and sessions, team agents, agent tokens, plan, review, message and confidence types, command rules, cost caps and confidence rules. The agents:run and models:* scopes are no longer offered by any preset, OAuth default or permission form, but a stored token or grant that names them still reads; and the removed operations are gone from the scope table, so the API's check that every operation has exactly one scope is strict again. What stored data needs in order to parse (old run and credential kinds, an agent team lead, retired rules and webhook events, old guardrail and integration settings) still parses, and a run's token is still PrincipalKind::Agent, which now means a run's token and the @g1t rows written before. The customize, marketplace, skills and claude-plugin contracts stay only until the mobile app drops its Customize screens.4d ago

subscribers.rsMerge branch 'worktree-agent-a4babc5b5815f6f3a'3d ago

teams.rsThe contracts no longer describe g1t's own agents, and workspaces can keep their AI providers. Every consumer had already stopped using them, so the TypeScript and Rust contracts drop the workspace-agent, ability, look, skill-library, context, runner-agent and mention modules, the chat and docs RPCs agents posted and read through, the folio agent methods, the agent_waiting notification, agent.asked and memory.changed events, AI credit, token and gateway billing types, model routes and sessions, team agents, agent tokens, plan, review, message and confidence types, command rules, cost caps and confidence rules. The agents:run and models:* scopes are no longer offered by any preset, OAuth default or permission form, but a stored token or grant that names them still reads; and the removed operations are gone from the scope table, so the API's check that every operation has exactly one scope is strict again. What stored data needs in order to parse (old run and credential kinds, an agent team lead, retired rules and webhook events, old guardrail and integration settings) still parses, and a run's token is still PrincipalKind::Agent, which now means a run's token and the @g1t rows written before. The customize, marketplace, skills and claude-plugin contracts stay only until the mobile app drops its Customize screens.4d ago

time.rsFast pages, required checks on the branch, self-hosted runners, honest incidents1w ago

tokens.rsAn access token can be regenerated: a new secret, its name, permissions, repositories and reach kept, and the old secret refused at once. Its page (yours under Settings, a workspace's under its settings for owners) has Regenerate with a new expiry, within the workspace's rules on lifetime, and shows the secret once; identity's regenerate_token logs it to the security log or the workspace's audit log, and refuses a revoked token. The token form's quick permissions have All, every resource at its highest level. A project's Observability settings no longer list the token g1t gives its apps on g1t.page: it is g1t's, and the page says apps there need none.4d ago

updates.rsTeams and CODEOWNERS, labels and milestones, dependency updates, the security suite, and a clearer top bar1w ago

webhooks.rsg1t coordinates the work and the person's own sessions do it: Ask g1t reads g1t with the same tools any agent gets, answers in one conversation shared with the @g1t direct message in Chat, and hands work to sessions in the person's own harness, whose subagents g1t records, ties to issues and pull requests, and lets see each other.4d ago

work.rsAgents in any harness coordinate through one inbox, the merge queue lands only what an attempt itself tested, and workspaces have Insights. Each session keeps the coordination updates that concern it (session_updates, migration 0038): claims, releases, hand-offs and expiries, overlaps, dependency changes, decisions, integration results and sessions starting and ending, about the issues and pull requests it is on or tied to, or every one while it is on nothing yet; never its own start, end or claims. An update is written in the same batch as the change it announces, guarded so it exists only once the change is made, with an id derived from the change, so a retry or two requests at once deliver it once; a blocker clearing and queue results follow their change under a key that names it. get_session's `updates` lists up to 100, oldest first, and reading never clears them; record_session_entries' `acknowledge` clears exact IDs for that session only and says how many were its (`acknowledged`). The CLI has `g1t session updates` and `g1t session acknowledge ID --note` (the decision note and the acknowledgement in one idempotent call, `--session` over G1T_SESSION), and those and `heartbeat` say which files the working copy changed (git status; G1T_TOUCH=0 turns it off), so overlapping edits are noticed whichever harness makes them; a pair of sessions is told again when another file in common joins. The Claude Code hooks fetch the inbox in the background, show it at every prompt and before tools when it changes or every two minutes, refresh after an acknowledgement, and never acknowledge. The AGENTS.md snippet, the skills, MCP's instructions and llms.txt say the same steps for every harness: scope claims (Scope, Interfaces, Depends on, Checks), proceed on complementary work, agree ownership before overlapping edits, record public summaries and evidence, never private reasoning. Claims resolve ownership in one statement, so simultaneous claims by one session converge. The merge queue: every attempt has its own id and branch (0037); a claimed build, an accepted build, a report and a callback each check their attempt; a dispatch registers its runs first and Actions fails the dispatch when it cannot; an attempt is decided only by its own runs (and other tools' statuses set after it was claimed), never an earlier attempt's or the pull request's own on the same commit (0040 keeps what it built and how many runs it registered); landing re-reads the rules, honours a bypass asked for when queued (0039), checks no predecessor remains, carries the exact tested commit to Git's compare-and-swap, records publication intent that fences removal, closing, drafting and changing base, resumes after a crash, and a publisher that lost a race no longer undoes the winner's landing; a retried attempt's branch is deleted; an entry whose branch or pull request is gone leaves the queue with a note instead of blocking those behind it; diagnosis rebuilds exactly what was tested and names a failing combination, not a proven cause. A session's page shows what awaits its agent as the coordination feed does. Workspace Insights (/:owner/-/insights): delivery metrics across a workspace's readable repositories (work service delivery_insights). Deploy Work migrations 0037 to 0040 first (the pipeline does), then Repos and Actions before Work, and the site with Work. Checked: cargo test for g1t-work (143), g1t-repos (311), g1t-actions-service (108), g1t-api (201), g1t (149), g1t-rules (72), g1t-contracts (306) and wasm32 builds; cargo check --workspace; the conformance suite against freshly built API and Work Workers (5, none skipped); the Claude Code hooks (49); the site's typecheck, tests (919) and a production build; the deploy tool's 92 tests; the docs build; and two unscripted local rehearsals in which two concurrent agents, given only a task and AGENTS.md, found each other, agreed a split and made complementary edits.2d ago