Claudeagent and ClaudeagentA signed archive link counts by the link, not by the address it comes from. `actions/checkout` in a job container without git downloads a tarball through the link `GET /repos/{owner}/{repo}/tarball/{ref}` sends it to; counted against the signed-out limit of 30 archive downloads a minute per address, jobs behind one address, as g1t's own machines are, would have used it up between them. A link is made for one commit, for 10 minutes, by a request the API already counts, so it is counted at the token rate instead. The rate limits reference says so.4f0f25fHistory
- ..
- about.rs
- archive_links.rs
- backups.rs
- blame.rs
- catch_up.rs
- coalesce.rs
- commit_file.rs
- contributors.rs
- diff.rs
- drift.rs
- fallback.rs
- forks.rs
- git_http.rs
- git_ops.rs
- hidden_refs.rs
- import.rs
- land.rs
- languages.rs
- last_commits.rs
- lib.rs
- license.rs
- lifecycle.rs
- limits.rs
- listing.rs
- meters.rs
- mirror.rs
- moves.rs
- namespaces.rs
- pack_cache.rs
- pack_limits.rs
- push_checks.rs
- push_commits.rs
- refs.rs
- refs_cache.rs
- registry.rs
- resilience.rs
- rule_facts.rs
- rules.rs
- run_access.rs
- secret_scan.rs
- shards.rs
- shared.rs
- signatures.rs
- stats.rs
- store.rs
- test_merge.rs
- transfer.rs
- workflow_gate.rs