Activity

1. harlan-github-agent

   added

   routine:daily-checkin

   on Oct 9, 2026
2. harlan-github-agent commented on Oct 9, 2026

   

   harlan-github-agentbot

   on Oct 9, 2026 – with Harlan GitHub Agent

   ContributorAuthor

   More actions

   [ACTION NEEDED] Daily check-in: 2026-10-09

   2026-10-09T20:40:00.000Z — daily-checkin on skilld-dev/skilld.dev | 2 found | 1 new | 1 already known | 0 outside allowed scope | 0 over 5 files | 1 issues requested

   AMBER skilld.dev

   Green deploys and production on main, but an out-of-band D1 migration import at 03:47Z threw 3,930 requests in the 04:00Z hour, and five discovery-ledger submissions are stalled forever on purpose holds.

   Check	Result	Detail
   skilld.report	Warn	Nightly health AMBER: 5 stalled submissions, 52 sync failures
   skilld.workers	Warn	3,946 threw; 3,930 in one hour during migration import
   skilld.database	Warn	Rejected decisions and churn only; migrations match at 0151
   sentry.skilld	Warn	74 unresolved; D1 timeout family recurred 15:45Z
   skilld.git / deploy / ci / home / skills	Pass	7a339c5a live, all three gates green
   skilld.baseline / analytics / signing-key	Pass	Fresh; 19 copies (13 run); key good 41 days

   Needs you

   1. Park purpose-held ledger submissions for review; they now read as stalled forever. Becomes the ledger-stall issue.
   2. Serve stale on failed cache recompute; the 04:00Z import outage is 20x the prior evidence. Re-proposes the stale-serve issue.
   3. Decide the migration window: 03:47Z is Melbourne daytime and cost 3,930 failed requests. Your call.

   Changed since yesterday

   • Recovered: GITHUB_TOKEN rotation landed, 89 days remaining
   • Recovered: purpose-404 refail fixed by fix(registry): stop purpose retries for missing sources #593, deployed 16:54Z
   • Recovered: tree-truncated refail stays quiet, fix(registry): complete large repository source refresh #575 holding
   • New: D1 import outage burst at 04:00Z, recovered
   • Improved: Unpinned skills answer a permanent 503 retry loop on skills-raw #577 (SKILLD-3W) quiet since Oct 8 20:09Z

   Unchanged work (3)

   • Escalate recurring D1 internal errors (Escalate recurring D1 internal errors to Cloudflare #393), timeout cluster at 15:45Z today
   • X spend policy (X discovery projects past the accepted $40 monthly spend line #179), $30.90/mo under the $40 line
   • Unpinned skills 503 retry loop (Unpinned skills answer a permanent 503 retry loop on skills-raw #577), quiet one more window

   Evidence

   Exit 1 | coverage complete | deployment 7a339c5a | baseline fresh, gap 35.7 h
   Archive: ~/.local/state/daily-checkin/skilld-dev/skilld.dev/2026-10-09T20-42-41-328Z-d397d5ae-94aa-45ae-95c8-a5a14d5e30df.json
   Pulse: 44 users, +300 skills/24 h, 21 repo changes, 19 copies (13 run, 6 install), 0 digests, $0 AI, X $30.90/mo Post-only (123/2000 today). Actual billing unavailable.

   A submission job that dead-letters on repository_purpose_review_required never updates its discovery_ledger row and reconcileLedger has no branch for that exception, so five rows (antfu/skills-pack, feitangyuan/onetake, trungtin/codex-delegate, tsogjavklann/awwwards-3d, aryanutkarsh/seecode) stay 'submitted' and alarm as stalled every night instead of parking for the review surface like the terminal-reason design intends.

   Target: shared/server/discovery-ledger.ts

   Verify with: pnpm vitest run test/unit/discovery-ledger-reconcile.test.ts

   Estimated to change 2 files.
3. harlan-zw

   closed this as not plannedon Oct 11, 2026