nobs-review（诺布斯评论）

Review

Multi-model review without runtime interception.

• A single shared contract per review: .reviews/<review_id>/contract.json
• Rounds accumulate in: .reviews/<review_id>/rounds/rNNN/

Interaction Order

1. Run initialization. Report detected development tool and confidence level.
2. Report risk analysis.
3. Report recommended reviewers.
4. Prompt the user to confirm or override selections.
5. Execute the review.

IF detected development tool confidence is low, THEN state the low confidence and proceed.

Workflow

1) Init

python3 skills/nobs-review/scripts/init_review.py --task "<what to review>"

To override tool detection:

python3 skills/nobs-review/scripts/init_review.py --task "<what to review>" --dev-tool claude

To refresh an existing review and preserve history:

python3 skills/nobs-review/scripts/init_review.py --id <review_id> --task "<updated scope>" --refresh

The init command performs the following actions:

• Detects the current development tool.
• Analyzes diff risk.
• Pre-enables recommended reviewers in the contract.
• Sets .reviews/.active_review_id.

2) Confirm Reviewers

Perform the following checks in .reviews/<id>/contract.json:

• reviewers[].enabled
• reviewers[].command_template

Accept the recommendation or select custom reviewers.

3) Run Round

python3 skills/nobs-review/scripts/run_review.py --id <review_id> --round auto

Or use the active ID:

python3 skills/nobs-review/scripts/run_review.py --round auto

To limit reviewers:

python3 skills/nobs-review/scripts/run_review.py --id <review_id> --round auto --reviewers codex_high,claude_opus

4) Merge Findings

python3 skills/nobs-review/scripts/merge_review.py --id <review_id> --round latest

The --round auto argument is treated as latest during the merge operation.

Outputs per round include: merged.md and resolution.md located in .reviews/<id>/rounds/rNNN/. The latest pointers at the review root (merged-latest.md, resolution-latest.md) are ephemeral. They are overwritten on each subsequent merge.

For persistent checklist edits, use the round-scoped resolution.md.

5) Review Auto-Fix Candidates

After merging, check auto-fix-candidates.md in the round directory or auto-fix-candidates-latest.md at the review root.

Findings listed in this file satisfy both conditions:

• fix_determinism = high (exactly one correct fix, no judgment required)
• fix_scope = local (change confined to 1–3 lines within a single function)

Perform the following steps for each fix:

• Apply the fix.
• Mark Applied: [x].
• Commit the changes.
• Run a new round to verify: python3 skills/nobs-review/scripts/run_review.py --round auto python3 skills/nobs-review/scripts/merge_review.py --round latest

Findings not appearing in auto-fix-candidates.md require human judgment and are tracked in resolution.md.

The auto_fix_threshold in contract.json → policy controls the boundary:

"auto_fix_threshold": { "determinism": "high", "scope": "local" }

To expand the set of auto-fix candidates, raise determinism to medium or scope to cross-file.

6) Fix and Re-review

python3 skills/nobs-review/scripts/run_review.py --id <review_id> --round auto
python3 skills/nobs-review/scripts/merge_review.py --id <review_id> --round latest

This process creates r001 -> r002 ->..., preserving history.

Rules

• IF a reviewer CLI is missing, THEN fail and do not fake the output.
• IF JSON output is invalid, THEN keep raw output and mark the parse as failed.
• No single round constitutes the final truth; comparison across multiple rounds is required.

Output Contract

• references/schema.md
• references/providers.md

Notes

• Deterministic logic resides in scripts; this skill is intentionally thin.
• For high-risk diffs, employ at least two strong reviewers.
• The user determines the final reviewer set; recommendations are advisory.