Torch Bisect Skill for Claude

PyTorch Git Bisect for TorchTitan

Bisect the PyTorch repo to find the commit that broke TorchTitan.

Always use absolute paths — shell state does not persist between Bash calls.

Phase 1: Gather Information

Record the TorchTitan directory (pwd). Then check if

~/.claude/torchtitan_torch_bisect_cache.json exists — if so, show cached

values and let the user reuse or override them.

Collect via AskUserQuestion (pre-fill from cache if available):

• PyTorch repo path — absolute path to local checkout
• Build command — shell command to recompile PyTorch after each checkout
• Test command — run from TorchTitan dir; must exit 0 on good commits,

non-zero on bad. Judge ONLY by exit code, never by log content.

• Good commit — a commit hash or a date (e.g., Feb 15)
• Timeout policy — if test exceeds 10 min, auto-mark bad or ask each time?

After collecting, offer to save answers to the cache file.

Phase 2: Setup

• Check GitHub connectivity: run curl -s -o /dev/null --connect-timeout 20 -w "%{http_code}" https://github.com.

If it fails or returns non-200, tell the user GitHub is not reachable and

suggest checking internet connectivity or proxy settings. On Meta internal

servers, suggest setting https_proxy=http://fwdproxy:8080. Do not proceed

until connectivity is confirmed.

• cd <pytorch> && git fetch origin (timeout 300000ms). On failure, ask user

to fix (remind about proxy/fwdproxy if it looks like a network error).

• Check git bisect log 2>&1 — if a bisect is active, ask user to reset or abort.
• Resolve the good commit:
• Hash: verify with git log -1 <hash>
• Date: git log origin/main --before="<DATE>T23:59:59" --format="%H %s" -1

— show result and confirm with user.

• Bad commit is origin/main. Show it to user.
• Show git rev-list --count <good>..origin/main and estimated steps (~log2).
• Run each bisect command individually so failures can be reported cleanly:
• git bisect start
• git bisect bad origin/main
• git bisect good <good>

If any step fails, show the error and run git bisect reset before asking

the user how to proceed.

Phase 3: Bisect Loop

Repeat until bisect finds the first bad commit:

A. Build — run build command in PyTorch dir (timeout 600000ms).

On failure: show last 50 lines, ask user to Skip / Retry / Abort.

B. Test — run test command in TorchTitan dir (timeout 600000ms).

Exit 0 = good, non-zero = bad. On timeout, follow user's policy from Phase 1.

Never analyze logs — use only exit code.

C. Record — git bisect good or git bisect bad in PyTorch dir.

• Output contains "is the first bad commit" → done, go to Phase 4.
• Output contains "Bisecting:" → continue loop.
• Cannot narrow further (too many skips) → tell user, suggest abort.

D. Progress — print step number, commit hash, good/bad, remaining count.

Phase 4: Report and Cleanup

• Capture git bisect log and git show --stat <bad_commit>.
• Extract PR number from commit subject ((#NNNNN) pattern). If found, run

gh pr view <N> --repo pytorch/pytorch --json title,body,url to get PR

details. Fallback URL: https://github.com/pytorch/pytorch/pull/<N>.

• git bisect reset.
• Present summary:

## Bisect Complete
**First bad commit:** <hash>
**Commit message:** <subject>
**Author:** <author> | **Date:** <date>

### Associated Pull Request
**PR:** #<N> — <title>
**Link:** https://github.com/pytorch/pytorch/pull/<N>
**Summary:** <first ~20 lines of PR body>

### Changed files
<git show --stat output>

### Full bisect log
<git bisect log output>

Omit the PR section if no PR number was found. Do NOT diagnose root cause.

Rules

• Always use absolute paths.
• Judge test results by exit code only — never analyze output.
• On unexpected errors, show the error and ask the user before proceeding.
• Keep the user informed after every bisect step.