prathamdby Fix PR feedback
Flags
| Flag | Default | Effect |
|---|---|---|
--pr <n|url> |
current branch PR | Target PR |
--no-push |
off | Keep commits local |
--no-reply |
off | Do not post review replies |
Missing values are BLOCKED.
1. Resolve and synchronize
Resolve owner, repo, number, URL, base, head branch, and remote head SHA. Block
on auth failure, missing/closed PR, dirty tree, or unsafe head checkout.
Fetch, check out head, and fast-forward to remote SHA; never reset or force.
Record: PR/head SHA | hunt counts | current finding | verdicts | commit/push | replies | terminal.
Done when local HEAD equals the PR head SHA and the ledger identifies the PR.
2. Hunt before editing
REQUIRED SUB-SKILL: Read ../gh/SKILL.md before any GitHub I/O. Hunt through
that skill: surfaces 1–2 with pr-threads.ts --json --open --complete; CI
snippets with ci-failures.ts --json --pr N --sha <head>. Surfaces:
- unresolved review threads, including outdated ones
- every comment page inside each thread
- review-comment API chains reconciled to thread roots
- actionable top-level review bodies
- actionable PR conversation comments
- PR CI on the head SHA: terminal non-success required/blocking checks and annotations
Every hunt reconciles review-comment chains unless every root is proved present.
Load recipe 3 in ./REFERENCE.md (REST reconcile); script exit is not proof
roots are present. Skip 4–5 only when --complete JSON moreReviews,
moreComments, and moreConvo are false. Recipe 6 always SHA-pins required or
blocking checks and annotations; ci-failures is drilldown only. Load remaining
recipes only after run exits. Record counts and page markers. No triage or
edit before all six passes finish. Normalize one finding per claim (source,
target, author, path/line, rule ID, body). Deduplicate identical keys from
./REFERENCE.md; preserve native reply targets. Done when pagination is
exhausted and every finding is in the ledger.
3. Triage every finding
Read surrounding code and trace the claimed path; reproduce when possible.
Assign one verdict: fix, reject, clarify, or already-fixed, with one
evidence line; record why if skipped. No edits until all findings have verdicts.
Done when all findings are triaged and rejects have concrete evidence.
4. Fix and verify
Apply only fix verdicts in focused edits. Run narrowest covering checks for
each fixed cluster; failed required checks are BLOCKED. Do not change code
for rejected or clarification findings. Done when every fix has a verified diff
or no code fix was needed.
5. Commit and push
When a diff exists, discard pre-drafted subjects. Read ../commit/SKILL.md and
run it with --unstaged so it drafts from the locked diff as
type: <concrete code action proved by dominant hunks>. Trailers default deny
per Commit clean-room in ./REFERENCE.md (identity trailers and harness
footers). Require no ban-list token and passing conversation-only test; reject
canonical excuses there. Skip commit on a clean tree. Unless --no-push, push
and verify remote SHA; never force. With --no-push, fixed findings become
AWAITING_PUSH. Remote movement or push rejection is BLOCKED. Re-read
git log -1 --format=%B; ban-list tokens or banned trailers are BLOCKED.
Apply Trailer hygiene in ../commit/REFERENCE.md if needed. Done when there is
no diff, or one verified clean-room commit is pushed.
6. Re-hunt until stable
Repeat all six passes after code or remote mutations; triage arrivals and repeat Steps 4–6. After actionable set changes, require two consecutive hunts with matching sets. Done when no finding is new or untriaged and stable hunts finish.
7. Reply and report
Unless --no-reply, skip targets whose replies satisfy the verdict, draft
remaining replies using ./REFERENCE.md, then apply
./references/unslop-reply-drafts.md. Preserve bot prefixes. Consolidate
shared targets. Post replies through loaded gh skill (pr-reply.ts); do not
invent raw gh. Re-hunt after replies; new findings return to Step 3. Before
retrying a failed reply, refetch its target; retry once. Second failure is
BLOCKED. Do not resolve threads unless asked. Report
source | finding | verdict | action | evidence, PR URL, commit/push state,
hunt counts, and unreplied items. Terminals: SUCCESS, NO_CODE_CHANGE,
AWAITING_PUSH, BLOCKED. Never SUCCESS with unreplied targets. After
interruption, restart at Step 1. Does not fix merge conflicts.