review-on-step
Assesses project progress against defined milestones and planning documentation.
Install
mkdir -p .claude/skills/review-on-step && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/13118" && unzip -o skill.zip -d .claude/skills/review-on-step && rm skill.zipInstalls to .claude/skills/review-on-step
Activation
This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.
Use when the user explicitly asks to review a named project phase, step, milestone, or similar planned slice of work. Resolve the project's planning docs, initialize missing planning docs if needed, then produce a reviewer-first assessment with findings, gaps, and verification coverage.Key capabilities
- →Normalize the target of a review request
- →Resolve project review context from AGENTS.md or CLAUDE.md
- →Initialize missing planning documents like IMPLEMENTATION_PLAN.md
- →Read relevant code and tests for the target
- →Evaluate the implementation for correctness and regression risk
- →Produce a reviewer-first assessment with findings and gaps
How it works
The skill normalizes the review target, resolves planning documents, initializes missing ones, reads context, evaluates implementation, and produces a reviewer assessment.
Inputs & outputs
When to use review-on-step
- →Reviewing project milestone progress
- →Assessing phase completion
- →Checking implementation gaps
About this skill
Review On Step
Overview
Use this skill only for explicit requests to review a named or numbered project step, phase, milestone, or equivalent planning unit.
This is a reviewer workflow, not a planning or implementation workflow.
Trigger
Activate this skill only when the user explicitly asks for a review of a step or phase, for example:
review step 3review phase 2review milestone 4code review for phase 4
Do not activate it for loose requests like:
what should we do nextplan phase 3implement step 2
Workflow
1. Normalize the target
Treat step, phase, milestone, and similar planning labels as the same target by default unless the project documents distinguish them materially.
If the project clearly uses multiple labels with conflicting meanings, stop and clarify with the user.
2. Resolve project review context
Always check for project instructions first:
AGENTS.mdCLAUDE.mdGEMINI.md.codex/project-instructions.md
Then resolve the core planning docs:
IMPLEMENTATION_PLAN.mdARCHITECTURE_NOTES.mdBACKLOG.md
For each core planning doc, use this path policy:
- Check the project root first.
- Check
docs/planning/second. - If neither exists, create it in
docs/planning/. - Do not relocate an existing root-level document.
This keeps existing projects stable while giving new projects a conventional home for planning docs.
If matching specs or plans exist under project planning folders, review them too. Check the existing project convention first. Typical locations include:
docs/planning/docs/specs/docs/superpowers/plans/docs/superpowers/specs/
3. Initialize missing planning docs when needed
If one or more core planning docs are missing, initialize them before continuing the review.
Create docs/planning/ if needed, then create only the missing files there with minimal starter structure.
Use these starter headings:
IMPLEMENTATION_PLAN.md# Implementation Plan## Summary## Current State## Phase Plan## Definition of Done
ARCHITECTURE_NOTES.md# Architecture Notes## System Overview## Responsibilities## Data Flow## Risks and Constraints
BACKLOG.md# Backlog## In Progress## Next Up## Later## Risks / Follow-Ups
State clearly when the review is proceeding against newly initialized planning docs instead of established project docs. Treat conclusions as lower-confidence in that case.
4. Read review context
Always read:
- any project instruction files that exist
- the resolved planning docs
- relevant code and tests for the requested target
- any matching spec or plan docs in the project's existing planning folders
5. Evaluate the implementation
Review for:
- correctness and regression risk
- architecture drift from intended responsibilities
- schema and persistence consistency
- ACL coverage
- operation ordering and idempotency guarantees
- Redis and Kafka responsibility boundaries
- missing tests and verification gaps
Adapt the architecture-specific checks to the project at hand. Do not assume Spring Boot, Redis, Kafka, or any specific stack unless the project docs or codebase establish them.
6. Produce reviewer output
Return:
- findings first, ordered by severity
- file references when possible
- open questions or assumptions
- a short summary of coverage and residual risk
Do not start implementation unless the user explicitly asks Codex to make changes.
Boundaries
Do not use this skill to:
- plan a new phase
- write an implementation plan
- start coding automatically
- infer a review request from approximate wording
When not to use it
- →When the user wants to plan a new phase
- →When the user wants to write an implementation plan
- →When the user wants to start coding automatically
Limitations
- →The skill does not infer a review request from approximate wording
- →The skill does not start implementation automatically
How it compares
This skill provides a structured, reviewer-focused audit of a specific project step, including initialization of missing planning docs, which differs from a general planning or implementation workflow.
Compared to similar skills
review-on-step side by side with the closest alternatives in the catalog.
| Skill | Installs | Updated | Safety | Difficulty |
|---|---|---|---|---|
| review-on-step (this skill) | 0 | 4mo | No flags | Intermediate |
| drift-analysis | 2 | 5mo | No flags | Intermediate |
| planning | 2 | 3mo | No flags | Intermediate |
| gsd-plan-checker | 1 | 4mo | No flags | Beginner |
Try saying
Example prompts that trigger this skill in your AI assistant.
You might also like
drift-analysis
avifenesh
This skill should be used when the user asks about "plan drift", "reality check", "comparing docs to code", "project state analysis", "roadmap alignment", "implementation gaps", or needs guidance on identifying discrepancies between documented plans and actual implementation state.
planning
citypaul
Planning work in small, known-good increments. Use when starting significant work or breaking down complex tasks.
gsd-plan-checker
toonight
Validates plans before execution to catch issues early
feature-planning
mhattingpete
Break down feature requests into detailed, implementable plans with clear tasks. Use when user requests a new feature, enhancement, or complex change.
linear-ticket
useautumn
Refine rough engineering thoughts into structured Linear tickets with GitHub permalinks
design-to-issues
cloverich
Creates a GitHub epic and child issues from a design document's implementation plan. Use when a design doc has a reviewed plan section ready to be converted into trackable work.