npx claudepluginhub mission42-ai/m42-claude-plugins --plugin m42-signs[--approve-all-high] [--reject-all-low]sonnet/reviewRuns Codex code review on local git state (working tree or vs base branch). Supports --wait/--background, --base <ref>, --scope auto|working-tree|branch.
/reviewReviews staged changes or recent commits across five axes—correctness, readability, architecture, security, performance—producing categorized findings with file:line references and fixes.
/reviewReviews HTML file for design anti-patterns, principles violations, and accessibility issues. Generates markdown report with status tables and recommendations.
/reviewReviews specified code scope via four specialists (quality, security, performance, architecture), producing summary, detailed findings, refactoring suggestions, prioritized action plan.
/reviewPerforms expert multi-LLM code review with inline PR comments on staged changes, open PRs, working tree, or paths. Checks LLM providers and queries focus areas.
/reviewDispatches the reviewer agent to review current branch code changes against code quality principles.
Interactively review pending learnings in the backlog. For each learning, you can Approve, Reject, Edit, Skip, or Quit.
--approve-all-high: Auto-approve all high-confidence pending learnings (with confirmation)--reject-all-low: Auto-reject all low-confidence pending learnings (with confirmation)test -f .claude/learnings/backlog.yaml && echo "EXISTS" || echo "NOT_EXISTS"Parse $ARGUMENTS for batch operation flags:
--approve-all-high: Batch approve high confidence learnings--reject-all-low: Batch reject low confidence learningsIf neither flag is present, run interactive review.
Based on preflight:
EXISTS: Proceed with review flowNOT_EXISTS: Show helpful message about empty backlogIf backlog doesn't exist (NOT_EXISTS from preflight):
## Review Learnings
No backlog found at .claude/learnings/backlog.yaml
The learning backlog is empty. You can:
- Add a learning manually: /m42-signs:add
- Extract from a session: /m42-signs:extract <session-id>
If backlog exists:
.claude/learnings/backlog.yamlstatus: pendingIf --approve-all-high flag is present:
confidence: highpending to approved for all matching learningsIf --reject-all-low flag is present:
confidence: lowpending to rejected for all matching learningsFor each pending learning (1 of N):
Show the learning in this format:
## Learning: <id> (<current>/<total>)
**Title**: <title>
**Confidence**: <confidence>
**Target**: <target>
### Problem
<problem description - multi-line>
### Solution
<solution description - multi-line>
### Source
- Tool: <source.tool>
- Command: <source.command>
- Error: <source.error>
---
Actions: [A]pprove | [R]eject | [E]dit | [S]kip | [Q]uit
Use AskUserQuestion with these options:
On Approve:
pending to approvedOn Reject:
pending to rejectedOn Edit:
On Skip:
pendingOn Quit:
When user selects Edit:
Present current values and ask which field to edit:
## Edit Learning: <id>
Current values:
- Title: <current title>
- Problem: <current problem>
- Solution: <current solution>
- Target: <current target>
Which field would you like to edit?
Use AskUserQuestion with options:
For the selected field:
After any edit:
"${CLAUDE_PLUGIN_ROOT}/scripts/validate-backlog.sh" .claude/learnings/backlog.yamlIf the learning was from automated extraction (source.tool != "manual"):
After edit mode completes ("Done Editing"):
When all pending learnings have been processed:
## Review Complete
| Action | Count |
|--------|-------|
| Approved | N |
| Rejected | N |
| Skipped | N |
Next steps:
- Run `/m42-signs:apply` to write approved learnings to CLAUDE.md files
- Run `/m42-signs:status` to see current backlog status