chore(skills): make /ship single-mode with plan/code ChatGPT review loops - #609
Merged
Conversation
…oops Collapse the attended/unattended split: /ship is now one autonomous flow — coordinator + fresh worker per unit, one integration branch and one PR per run. Every unit's plan iterates through a coordinator-run ChatGPT plan review loop (exit on VERDICT: APPROVED, max 5 passes); the PR iterates the existing code review loop (exit on VERDICT: SHIP, max 3, script-enforced). Auto-merge without prompting when all proof conditions hold at one exact head; any failure — including loop exhaustion — is a full stop for a human decision. Also fixes the inconsistencies found in review: the local-e2e contradiction in unattended.md, the unsatisfiable "CI all engines" claim, the stale "third pass" merge rule in ship-phase (deleted) and both wiki pages, the steps-2–5 wording drift, the Chrome single-session concurrency hazard (all chatgpt-review invocations are coordinator-serialized), and vague "approved temporary directory" phrasing. per-issue-cycle.md is renumbered to its own steps 1–4. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01E4bdx3RwYDjVdPdFCggKVQ
This was referenced Aug 5, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Restructures the
/shipskill into a single autonomous mode (the former unattended flow) and adds two explicit ChatGPT review loops:chatgpt-reviewplan mode, exit onVERDICT: APPROVED, max 5 passes; exhaustion → full stop, human decides before any code is written.VERDICT: SHIPat the exact head with no accepted findings, max 3 passes (script-enforced); certification + green CI + branch protection → auto-merge with no prompt. Any failed proof condition — including 3 passes without certification — → full stop, human decision needed (replaces silent "leave PR open").Architecture kept: coordinator never implements; fresh worker subagent per unit bounds context; one integration branch and one PR per run.
Consistency fixes folded in
unattended.mdclaimed Playwright doesn't run locally while the worker contract mandates it) — workers run Chromium+WebKit locally; PR CI adds Chromium e2e (Playwright e2e never runs on PRs or main — green CI does not mean the browser suite passed #564).ship-phase(its redirect note stated a wrong merge rule: "clean third pass" instead of first-clean-pass certification); same stale claim fixed in.wiki/Project-Skills.mdand.wiki/Development-Workflow.md.chatgpt-reviewinvocations are coordinator-run and serialized; workers never invoke the skill (new footguns section, incl. plan-session identity = plan-file path).per-issue-cycle.mdrenumbered to its own steps 1–4; worker-contract wording drift ("steps 2–5" vs "2–4 + CHANGELOG") resolved;$TMPDIRnamed explicitly.Process/tooling only — no
src/or product behavior changes, no CHANGELOG entry.🤖 Generated with Claude Code
https://claude.ai/code/session_01E4bdx3RwYDjVdPdFCggKVQ