Repository navigation
(standardize-repo): Report each repo's devcontainer definition, CI image and staleness, and queue rebuilds #1670
Description
Problem
Running containers go stale after a template update: the maintainer rarely rebuilds, and the staleness check is warn-only and not surfaced at session start. A standardize run should verify, per repo, that the devcontainer definition rendered, that the devcontainer builds and devcontainer-verify passed on the PR head, that the ruleset requires devcontainer-verify (seven managed repos did not on 2026-10-26), and after merge that the devcontainer-build run on the default branch for the merged head succeeded and pushed (read from the workflow run, no packages scope), so the GHCR image matches the definition. It should list in the fleet report a rebuild queue, one command per workspace (devcontainer up for the CLI path, the Coder rebuild for Coder workspaces, or task devcontainer:rebuild once harmon-init#1833 ships), ordered after a wave's merges. Before any lane starts, the orchestrator's own container must pass the staleness probe and meet the release's tool floors; lanes in a stale container fail the new gates and look like template bugs. It should never rebuild inside a run: an agent cannot rebuild the container it runs in, a rebuild mid-wave kills every lane in the workspace, and the hard rules reserve process termination for the maintainer.
Acceptance criteria
- [CI] The fleet run checks the orchestrator's own container against the staleness probe and the release's tool floors before creating the epic, and stops with the rebuild command when it fails.
- [CI] The per-repo report has a devcontainer row: definition rendered, devcontainer builds and
devcontainer-verifyon the head, whether the ruleset requiresdevcontainer-verify, the default-branch devcontainer-build run for the merged head (succeeded and pushed), and whether the staleness probe can run. - [CI] The fleet report lists the rebuild queue with one command per workspace the decisions file names, ordered after a wave's merges.
- [CI]
SKILL.mdstates that a run never rebuilds a container.
Verify
Re-read the ADR and spec named under Provenance for the current design, and re-establish every path, task name or observation above with git grep in the target checkout (and gh for GitHub state) before acting; a renamed path means the claim needs re-checking, not that the problem is gone.
Out of scope
SessionStart staleness summary and host-side rebuild task (harmon-init).
Provenance
Filed from the 2026-10-26 standardize-repo redesign session (harmon-devkit, worktree/clear-river-0602). Design record: harmon-devkit ADR docs/decisions/2026-10-26-fleet-standardization-runs.md and spec specs/standardize-repo-fleet.md.
Activity
evanharmon1 commentedon Nov 2, 2026
Claiming — starting implementation on branch feat/1670-devcontainer-report (session fleet-m6-foundations).
Preflight (§3):
Claim record (for
/wrap— undo only what this claim added):claim:label added by this claim: claim:gptclaim:model label added by this claim: n/aclaim:label displaced by this claim: noneclaim:label owned by this claim chain: claim:gptclaim:model label owned by this claim chain: n/aclaim:label displaced by this claim chain: none10 remaining items
evanharmon1 commentedon Nov 3, 2026
Dev flow v2 evidence: review round 1 for
run-1670-devcontainer-reportat1cc912e(findercodex-verification).{ "run_id": "run-1670-devcontainer-report", "stage": "review", "round": 1, "reviewed_head": "1cc912e05798b70194b6916679d0774a2ccbf9e8", "passes": [ { "schema": 2, "role": "reviewer", "status": "completed", "head": "1cc912e05798b70194b6916679d0774a2ccbf9e8", "produced_at": "2026-11-03T03:13:56Z", "producer": { "harness": "codex-cli", "model": "gpt-6-astra", "tier": "frontier" }, "run": { "run_id": "run-1670-devcontainer-report", "initiated_by": "human" }, "payload": { "stage": "review", "round": 1, "reviewed_head": "1cc912e05798b70194b6916679d0774a2ccbf9e8", "finder": "codex-verification", "slot": "codex-verification", "findings": [ { "id": "review-r1-codex-verification-1", "path": "ai/skills/repo/standardize-repo/assets/fleet-report.py", "line": 239, "class": "correctness", "provenance": "original", "fingerprint": "new", "priority": "P2", "recommended_disposition": "fix", "evidence": "Codex [P2] 'Preserve significant whitespace in rebuild commands' (fleet-report.py:238-240): the rebuild command is rendered as ordinary Markdown table text via text(), so GitHub's rendered view collapses repeated spaces inside a quoted argument (e.g. devcontainer up --workspace-folder '/work/my app' copies as a different path); escaping Markdown punctuation alone does not deliver the 'verbatim' display references/fleet-report.md promises; asks for a whitespace-preserving representation plus a regression test. Reviewer check via git show 1cc912e0:ai/skills/repo/standardize-repo/assets/fleet-report.py: line 239 builds the cell with text(entry[\"command\"]); text() (lines 26-33) html-escapes, backslash-escapes \\`*_{}[]()|!#@ and replaces CR/LF, but keeps consecutive spaces as raw spaces, which HTML rendering collapses. Running text() on the example returns the double space unchanged in source. Confirmed as described." }, { "id": "review-r1-codex-verification-2", "path": "ai/skills/repo/standardize-repo/assets/devcontainer-probe.sh", "line": 92, "class": "correctness", "provenance": "original", "fingerprint": "new", "priority": "P2", "recommended_disposition": "fix", "evidence": "Codex [P2] 'Retain other known workflow-run conclusions' (devcontainer-probe.sh:91-93): when the merged-head devcontainer-build run completes cancelled, timed_out or action_required, the API returns a definite conclusion, but the whitelist reports 'unknown', making a known unsuccessful run indistinguishable from missing evidence; asks to preserve those conclusions (push still unknown) and test them. Reviewer check via git show 1cc912e0:ai/skills/repo/standardize-repo/assets/devcontainer-probe.sh: line 92 is `result = conclusion if conclusion in ('success', 'failure', 'skipped') else 'unknown'`, so any other completed conclusion (cancelled, timed_out, action_required, neutral, stale, startup_failure) is mapped to 'unknown'; test-devcontainer-probe.sh test_image_conclusions covers only failure/skipped/missing-run. Confirmed as described." } ], "counts": { "P0": 0, "P1": 0, "P2": 2, "P3": 0 } } } ], "adjudication": { "schema": 2, "run_id": "run-1670-devcontainer-report", "stage": "review", "round": 1, "reviewed_head": "1cc912e05798b70194b6916679d0774a2ccbf9e8", "adjudications": [ { "finding_id": "review-r1-codex-verification-1", "reviewer_priority": "P2", "adjudicated_priority": "P2", "disposition": "fix", "reason": "Confirmed by reading: rebuild commands render through text() into a plain Markdown table cell, so GitHub collapses repeated spaces, and $…$ / ~…~ can render as math or strikethrough; the copied command differs from the decisions file. Render each command as a code span whose backtick fence is longer than any backtick run inside, with pipes escaped for the table and padding where it starts or ends with a backtick or space, and test it.", "evidence": "fleet-report.py@1cc912e0:239; text() line 33", "override": null }, { "finding_id": "review-r1-codex-verification-2", "reviewer_priority": "P2", "adjudicated_priority": "P2", "disposition": "fix", "reason": "Confirmed by reading: only success, failure and skipped pass through; a completed run concluding cancelled, timed_out, action_required, neutral, stale or startup_failure reads as unknown, hiding an observed result. Report any completed conclusion from that documented set verbatim, keep unknown for a missing, in-progress or unrecognised run, and test cancelled and timed_out.", "evidence": "devcontainer-probe.sh@1cc912e0:92", "override": null } ] }, "exit": { "stage": "review", "rounds_counted": 1, "next_round": 2, "substitutions": [], "split_candidate": { "detected": false, "round": 1, "mechanism": null, "concentration": 0, "provenance_share": 0, "introduced_by_rounds": [], "finding_ids": [], "consecutive_rounds": [ 1 ], "reason": "no_gating_findings" }, "outcome": "continue", "reason": "below_threshold", "action": "dispatch", "corrections": [], "diagnostics": [], "verified_findings": [ { "id": "review-r1-codex-verification-1", "provenance_status": "verified", "verified_provenance": "original", "fingerprint_status": "verified", "verified_fingerprint": "new" }, { "id": "review-r1-codex-verification-2", "provenance_status": "verified", "verified_provenance": "original", "fingerprint_status": "verified", "verified_fingerprint": "new" } ] } }Round
review round 1 — reviewed 1cc912e
…can render as math or strikethrough; the copied command differs from the decisions file. Render each command as a code span whose backtick fence is longer than any backtick run inside, with pipes escaped for the table and padding where it starts or ends with a backtick or space, and test it.Exit: continue — below_threshold (rounds counted: 1; next round: 2)
fix(standardize-repo): render rebuild commands as code spans and repo…
evanharmon1 commentedon Nov 3, 2026
Dev flow v2 evidence: review round 2 for
run-1670-devcontainer-reportat800cf34(findercodex-verification).{ "run_id": "run-1670-devcontainer-report", "stage": "review", "round": 2, "reviewed_head": "800cf34ba113ad96b48a0b9134f7140131abf0eb", "passes": [ { "schema": 2, "role": "reviewer", "status": "completed", "head": "800cf34ba113ad96b48a0b9134f7140131abf0eb", "produced_at": "2026-11-03T13:58:49Z", "producer": { "harness": "codex-cli", "model": "gpt-6-astra", "tier": "frontier" }, "run": { "run_id": "run-1670-devcontainer-report", "initiated_by": "human" }, "payload": { "stage": "review", "round": 2, "reviewed_head": "800cf34ba113ad96b48a0b9134f7140131abf0eb", "finder": "codex-verification", "slot": "codex-verification", "findings": [], "counts": { "P0": 0, "P1": 0, "P2": 0, "P3": 0 } } } ], "adjudication": { "schema": 2, "run_id": "run-1670-devcontainer-report", "stage": "review", "round": 2, "reviewed_head": "800cf34ba113ad96b48a0b9134f7140131abf0eb", "adjudications": [] } }feat(standardize-repo): report devcontainer status, CI image and rebu…
github-actions commentedon Nov 3, 2026
Claim released — PR #1823 merged (claiming work unit complete). (Supersedes the claim record above.)
Released by claim-release automation:
claim:gptlabel: removedevanharmon1: removedgithub-actions commentedon Nov 3, 2026
Claim released — issue closed (completed). (Supersedes the claim record above.)
Released by claim-release automation:
claim:gptlabel: removedevanharmon1: removedevanharmon1 commentedon Nov 3, 2026
Slate membership:
evanharmon1 commentedon Nov 3, 2026
Slate membership: