From pro-plan
Review and revise implementation plans: tech-stack-aware best practices, DRY checks, structure analysis, and completeness review.
How this skill is triggered — by the user, by Claude, or both
Slash command
/pro-plan:pro-planThis skill is limited to the following tools:
The summary Claude sees in its skill listing — used to decide when to auto-load this skill
Review an existing implementation plan or TDD specified in $ARGUMENTS. If no path is given, find the most recent plan from the Claude Code plans directory.
Review an existing implementation plan or TDD specified in $ARGUMENTS. If no path is given, find the most recent plan from the Claude Code plans directory.
Supports two modes:
--review-only: review and ask questions, but skip revision.--revise-only: restructure and improve the plan without running the review lenses.Extract from $ARGUMENTS:
ls -t .claude/plans/*.md 2>/dev/null | head -1
If no plans directory or no files found, error: "No plan file found. Pass a path: /pro-plan path/to/plan.md"--review-only → run Phase 1 only (review + questions, no revision)--revise-only → run Phase 2 only (revise without review)If both flags are set, error: "--review-only and --revise-only are mutually exclusive."
Read the plan file. If it doesn't exist, error with the path that was tried.
Count lines to classify plan size:
Scan the plan text for technology mentions. Look for:
Also check the repo for stack signals:
ls package.json tsconfig.json go.mod go.sum Cargo.toml pyproject.toml requirements.txt Gemfile pom.xml build.gradle docker-compose.yml Dockerfile Makefile 2>/dev/null
Combine plan mentions and repo signals into a deduplicated tech stack list.
For each detected technology (cap at 5 technologies), run a WebSearch for authoritative best-practices resources. Construct queries like: "[technology] best practices", "[technology] design patterns", "[technology] common pitfalls".
Cap total web searches at 8. For each search, briefly note the top authoritative result (title, source, key takeaway).
Resource priority (prefer in this order):
Store the collected resource summaries — these are passed to all agents as source-of-truth context.
ls CONTRIBUTING.md contributing.md CONTRIBUTING 2>/dev/null
If found, read it. This is a source of truth alongside the web-searched resources — it informs all agents' judgment about what the plan should and shouldn't do.
Read the shared agent files:
${CLAUDE_SKILL_DIR}/../../agents/shared/calibration.md${CLAUDE_SKILL_DIR}/../../agents/shared/taxonomy.md${CLAUDE_SKILL_DIR}/../../agents/shared/style.mdIf --revise-only: Skip to Step 10.
Otherwise: Continue to Step 8 (review phase).
Read the four lens agent files:
${CLAUDE_SKILL_DIR}/../../agents/lens-dry.md${CLAUDE_SKILL_DIR}/../../agents/lens-best-practices.md${CLAUDE_SKILL_DIR}/../../agents/lens-structure.md${CLAUDE_SKILL_DIR}/../../agents/lens-completeness.mdLaunch 4 Agent calls in parallel. Each agent receives:
Each lens produces structured findings: {section, label, body, severity, reasoning, lens}. Lenses must be independent — do not let one lens's output influence another.
Load the review synthesizer:
${CLAUDE_SKILL_DIR}/../../agents/review-synthesizer.mdFeed it all findings from all 4 lenses plus the original plan text.
Print the structured review to the terminal:
## Plan Review: [filename]
**Tech stack:** [detected technologies]
**Plan scope:** [Short/Medium/Long] ([N] lines, [M] phases)
**Resources consulted:** [list of resources used]
---
### Critical
> **[label]:** [body]
> *Section: [plan section]* | *Lens: [source]*
> **[Fix/Add/Mitigate]:** [direction]
### High
> ...
### Medium
> ...
---
Stats: [N] raw → [N] dedup → [N] posted ([breakdown by tier])
If the synthesizer produced questions, present them via AskUserQuestion. Cap at 5 questions. Group by theme.
If --review-only: Stop here. Print a summary line and exit.
Load the plan reviser:
${CLAUDE_SKILL_DIR}/../../agents/plan-reviser.mdFeed it:
--revise-only)Derive the output path from the original plan path:
my-plan.md → my-plan-pro.mddesign.md → design-pro.mdWrite the revised plan to this new file in the same directory as the original.
Print a one-line summary:
npx claudepluginhub asbarron/claude-skills --plugin pro-planOrchestrates parallel architecture and experience reviews of implementation plans, scores findings across dimensions like data flow and UX, consolidates ranked fixes for user approval and auto-application. Use after planning, before non-trivial coding.
Reviews implementation Plan files in parallel using Codex, Gemini, and Claude to analyze validity, gaps, risks, and improvements. Invoke via /plan-review after plan creation.
Reviews and validates plans (implementation, research, design, migration, etc.) using parallel subagents for codebase alignment, best practices, standards, feasibility, and fresh perspectives.