From arn-code
Validates structured Arness project plans for completeness, correctness, and pattern compliance before execution. Checks documents, dependencies, codebase references; classifies issues by severity, suggests fixes, and proceeds to taskification.
How this skill is triggered — by the user, by Claude, or both
Slash command
/arn-code:arn-code-review-planThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Validate and review a structured project plan for completeness, correctness, and pattern compliance before execution. Reports issues with severity classification and actionable fix suggestions, offers interactive remediation, and optionally proceeds to task creation.
Validate and review a structured project plan for completeness, correctness, and pattern compliance before execution. Reports issues with severity classification and actionable fix suggestions, offers interactive remediation, and optionally proceeds to task creation.
Read the project's CLAUDE.md and look for a ## Arness section. Extract:
If ## Arness is not found: inform the user: "Arness is not configured for this project yet. Run /arn-planning to get started — it will set everything up automatically." Do not proceed.
Ask the user for PROJECT_NAME if not provided in the conversation.
Read all project files from <plans-dir>/<PROJECT_NAME>/:
<plans-dir>/<PROJECT_NAME>/
├── SOURCE_PLAN.md
├── INTRODUCTION.md
├── TASKS.md
├── PROGRESS_TRACKER.json
└── plans/PHASE_*.md
If any critical files are missing (SOURCE_PLAN.md, INTRODUCTION.md, TASKS.md, or any PHASE_*.md), report the missing files immediately. Do not attempt to run checks against missing files — instead, note the missing file as an ERROR finding and continue checking the files that do exist.
Read ${CLAUDE_PLUGIN_ROOT}/skills/arn-code-review-plan/references/validation-checks.md and execute all generic checks (sections 1-4, 6-8) against the project files.
Classify each finding by severity:
For each finding, record:
Suggested fixes must be specific and actionable. Examples:
reports/IMPLEMENTATION_REPORT_PHASE_2.json"Validate that references in the plan actually exist in the codebase:
Add findings for any broken references. Suggestions should be specific:
src/old/module.py to src/new/module.py in Codebase References table"conftest.py:db_session — fixture no longer exists"src/services/base.py:15-30 — update snippet"Read stored pattern docs from the Code patterns path in the ## Arness config:
<code-patterns-dir>/code-patterns.md<code-patterns-dir>/testing-patterns.md<code-patterns-dir>/ui-patterns.md (if it exists)Code pattern compliance:
For each pattern documented in code-patterns.md:
Testing pattern compliance:
For each testing pattern in testing-patterns.md:
UI pattern compliance (if ui-patterns.md exists):
For each pattern documented in ui-patterns.md:
ui-patterns.md documents themAll pattern compliance findings are WARNING severity (not ERROR), since pattern compliance is advisory — the patterns may have been intentionally overridden for a specific phase.
Each finding includes a suggestion, e.g.:
@pytest.mark.unit marker per testing patterns"Present a human-readable summary to the user:
## Plan Review: <PROJECT_NAME>
### Summary
- Phases: N | Implementation tasks: N | Testing tasks: N
- Issues: N errors, N warnings, N info
### ERRORS (must fix)
1. [S001] INTRODUCTION.md missing — <project-dir>/
→ Fix: Re-run arn-code-save-plan to regenerate INTRODUCTION.md
2. [T013] Task 5 depends on "Task 9" but only 6 tasks exist — TASKS.md:45
→ Fix: Update dependency to correct task number
...
### WARNINGS (should fix)
1. [I012] Pattern "Service Layer" missing "How to apply" — INTRODUCTION.md:45
→ Fix: Add "How to apply" section with concrete instructions
...
### INFO (optional)
1. [Q004] Phase 3 has 9 implementation tasks — consider splitting
...
Every ERROR and WARNING includes a → Fix: suggestion line.
If no issues are found, report: "Plan passed all checks — no issues found."
Present the user with options based on what was found:
If ERRORS exist:
Ask the user (using AskUserQuestion):
If only WARNINGS exist (no errors):
Ask the user (using AskUserQuestion):
If no issues: skip this step entirely.
For each fix applied:
After remediation:
remediated: true and the action taken)Generate a JSON report using the REVIEW_REPORT_TEMPLATE.json from the configured template path. Populate all sections:
ready (0 errors), fix_errors (errors remain), or needs_review (warnings remain)arn-code-taskify if ready, or fix errors firstSave the report to: <plans-dir>/<PROJECT_NAME>/reports/REVIEW_REPORT.json
Present the final status to the user:
Offer the next step:
/arn-code-taskify to create the Claude Code task list?"/arn-code-taskify with the same PROJECT_NAME## Arness config is missing, do not proceed — suggest running /arn-planning to get started./arn-code-save-plan first./arn-planning to generate pattern docs on first use."npx claudepluginhub appsvortex/arness --plugin arn-codeVerifies post-execution implementation against project plan, code patterns, testing patterns, and architecture docs. Classifies issues as ERRORS, WARNINGS, INFO with verdict.
Validates implementation plans against codebase reality, architecture, quality, risks, and conventions before execution using four parallel specialized reviewers.
Validates AI implementation plans before execution. Use when user says "check your plan", "validate this plan", "review the plan", or "is this plan good". Launches 5 parallel validators + devil's advocate.