NodeJSmith/mine-review

Technical Review

Dispatches three parallel reviewers adapted to content type. Code: correctness, integration fit, readability. Instruction files: consistency, instruction quality, writing quality. Consolidates into a prioritized report.

Arguments

$ARGUMENTS — optional scope. Empty for full branch diff, or a directory/file list. Path mode when files exist with no branch changes.

Phase 1: Determine Scope

Read and execute ${CLAUDE_CONFIG_DIR:-~/.claude}/skills/mine-review/scope-detection.md.

Phase 2: Dispatch Reviewers

Pre-compute diff (diff mode only)

1. get-skill-tmpdir mine-review → <tmpdir>
2. git rev-parse HEAD → <sha>
3. <diff command> > <tmpdir>/diff.patch

Detect review mode

A file is an instruction file if it has a .md extension. If ALL files are instruction files, use instruction mode. Otherwise code mode.

Build scope line

• Diff mode: Diff at <tmpdir>/diff.patch (HEAD: <sha>). Read changed files for surrounding context. Before relying on the diff, verify HEAD matches <sha>.
• Path mode: Review these existing files (not a diff): <file list>
• Instruction diff mode — append: Read every changed file in full, not just diff hunks.

Code mode — dispatch all three in one message

Each agent has its own checklist and output format. Pass scope only:

• subagent_type: "code-reviewer" — Review changes. <scope line>
• subagent_type: "integration-reviewer" — Review changes. <scope line>
• subagent_type: "wtf-reviewer" — Review changes. <scope line>

Instruction mode — dispatch all three in one message

• subagent_type: "fine-toothed-comb" — Review these instruction files for consistency, accuracy, and completeness. <scope line>. Read sibling files for cross-reference context.
• subagent_type: "instruction-quality-reviewer" — Review these instruction files. <scope line>
• subagent_type: "writing-quality-reviewer" — Review these instruction files. <scope line>

Phase 3: Consolidate and Present

Deduplicate

If two reviewers flagged the same issue, keep one entry and note the cross-signal: (flagged by code-review + readability pass).

Validity assessment

Apply the protocol from ${CLAUDE_CONFIG_DIR:-~/.claude}/skills/mine-challenge/findings-protocol.md: findings valid by default; flagging one as likely invalid requires a concrete evidence trail.

Present the report

Organize by severity, not by reviewer. Report header:

## [Technical|Instruction] Review: [branch name or target path]

**Scope:** N files changed, +X/-Y lines (diff) | N files, X total lines (path)
**[Reviewer 1]:** PASS / WARN / FAIL
**[Reviewer 2]:** PASS / WARN / FAIL
**[Reviewer 3]:** X findings (N HIGH, N MEDIUM, N LOW)
**Likely-invalid:** N

Source labels — code mode: Code, Integration, Readability. Instruction mode: Consistency, Instruction, Writing.

Findings in severity-grouped tables (### Critical / High, ### Medium, ### Low). Each finding includes the proposed fix — the specific edit that would be applied. The user decides what to fix based on seeing both the problem and the proposed edit. Likely-invalid findings in a separate section with Claimed/Actually/Why-invalid fields.

Next steps

AskUserQuestion:
  question: "What would you like to do with these findings?"
  header: "Next steps"
  multiSelect: false
  options:
    - label: "Fix all"
      description: "Apply the proposed fixes listed above, highest severity down, then re-read the modified content"
    - label: "Fix critical/high only"
      description: "Address blockers, leave medium/low for later"
    - label: "Note and move on"
      description: "Acknowledged — no fixes this session"

If fixing: work through findings top-down by severity, making edits directly. After fixes: "Fixes complete — run /mine-commit-push or proceed to commit when ready."

What This Skill Does NOT Do

• Deep codebase audit — use /mine-audit
• Exhaustive style sweep — use /mine-clean-code
• Automatic fixing without asking — it diagnoses, then asks what to fix