docs(ship): one ChatGPT session per unit, start to finish - #647
Merged
Conversation
During /ship 630, when a unit's code-review loop hit its 3-pass cap with no certified head, the coordinator opened a fresh ChatGPT session for a "single-shot certification check" rather than continuing the existing conversation. The user corrected this explicitly: use a single session per unit across plan authoring, every review round, and any ad hoc advisory consultation — never session: null a second time. Encode this as a hard rule in review-loops.md, including the concrete fact discovered while applying it: chatgpt-review.mjs's pr mode hard- rejects a 4th call tied to one session (a tool-level cap, not just the workflow script's loop bound), so continuing past pass 3 means driving the existing conversation tab directly rather than re-invoking the workflow with a higher pass number. Cross-reference from SKILL.md's merge-gate FULL STOP section, the exact point where the mistake happened. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
What & why
During
/ship 630(issue #630, phase 5), when a unit's code-review loop hit its 3-passcap with no certified head, the coordinator opened a fresh ChatGPT session for a
"single-shot certification check" rather than continuing the existing conversation that
had already reviewed the plan and every prior pass. The user corrected this explicitly
mid-run: use a single ChatGPT session per unit, from plan authoring through merge,
including any ad hoc advisory consultations — never open a second session to route
around an exhausted pass cap.
This encodes that correction as a hard rule in
skills/ship/references/review-loops.md(under "Hard rules (all loops)"), including a concrete fact discovered while applying
it live:
chatgpt-review.mjs'sprmode hard-rejects a 4th call tied to one session(
"PR review sessions permit at most three total passes", a tool-level cap enforced bythe CLI itself, not just the workflow script's own loop bound) — so continuing past
pass 3 means driving the existing conversation tab directly (DOM read/insertText/click,
same technique already used for the Chrome-crash recovery path) rather than re-invoking
the workflow with a higher pass number. Cross-referenced from
SKILL.md's merge-gateFULL STOP section, the exact point in the process where the mistake happened.
Not part of any #630 phase — a process/tooling correction to the
/shipskill itself.Test plan