Skip to main content
← Back to list
01Issue
FeatureOpenSwamp CLIPublic
AssigneesNone

Relationships

#3514 swamp-extensions CI: Review Integrity explains a stalled or failed audit (streamed progress, timeout, did-not-finish comment)

Opened by skunk-ape · 10/15/2026

Follow-up to #3174, whose cause turned out to be runner pods evicted for exceeding their ephemeral-storage limit (fixed in infra and by shrinking the job). The in-step design built and tested for #3174 was deferred out of that fix, at the maintainer's call.

Land it in the Audit trust-root changes step of .forgejo/workflows/ci.yml, as described in the parked comment on #3174:

  • stream claude's events (stream-json, verbose) through a filter that prints typed fields only and redacts the key;
  • bound the call with a timeout below Forgejo's stale-task window (e.g. timeout --kill-after=15s 480);
  • cap retries (CLAUDE_CODE_MAX_RETRIES, API_TIMEOUT_MS), isolate CLAUDE_CONFIG_DIR, disable the autoupdater;
  • on timeout, post fixed did-not-finish text.

A real 429 or a hung API would then explain itself in the log and on the PR. Separately, decide whether a follow-up job with needs: review-integrity and if: failure() can post a did-not-complete comment when the runner itself dies; confirm first that Forgejo schedules it after a reaped job.

Keep the audit fail-closed; no partial audits.

02Bog Flow
◉OPEN○TRIAGED○IN PROGRESS○SHIPPED

Open

10/15/2026, 8:51:23 PM

No activity in this phase yet.

03Sludge Pulse

Sign in to post a ripple.