Review Approach
SkillMediaPerforms the final review of a completed proposal, product specification, technical design, and test plan for cross-artifact consistency, consequential gaps, missing decisions, weak tradeoffs, failure modes, testing strategy, and better alternatives; use when asked to review an approach, challenge design or testing decisions, find gaps in completed definition artifacts, or provide a second opinion before task planning or implementation.
Available today. Use it from your connected AI after setup.
No other account needed.
Connect ahel once, and every AI you use reads what you have installed.
Then ask your AI: use the Review Approach skill
What this skill tells your AI
The instructions your AI receives, as published by codagent-ai/agent-skills in skills/review-approach/SKILL.md and read by ahel’s review.
Review the completed proposal, specifications, design, and test plan together before task planning. Determine whether they form a coherent, implementation-ready definition with sound behavioral, technical, and testing decisions.
Read every definition artifact and relevant repository instructions. Inspect affected code, tests, and established patterns enough to judge feasibility and fit. Treat omissions as findings when an implementer or tester would otherwise have to invent a consequential decision.
Review focus
Trace important user flows and system interactions end to end. Look for material:
- contradictions, scope drift, dropped commitments, terminology drift, or incompatible assumptions;
- missing or untestable behavior, boundaries, state transitions, failure handling, permissions, data lifecycle, compatibility, or migration decisions;
- unclear ownership, component boundaries, data flow, concurrency, recovery, security, operability, rollout, or observability;
- testing strategy that conflicts with the definition, misses important integration or critical journeys, uses the wrong test layer, leaves acceptance substitutes ambiguous, or relies on unstated human judgment;
- avoidable coupling, complexity, maintenance burden, irreversible commitments, weak rationale, or overlooked alternatives.
Apply judgment rather than mechanically filling a checklist. Do not review task decomposition, implementation code quality, formatting, or parser mechanics except where they prevent the definition from being usable. Do not relitigate whether the approved feature should exist, broaden its scope, or invent optional features.
Report
For each consequential finding, cite the exact artifact section and repository evidence, classify it as a cross-artifact inconsistency, missing decision, or challenged decision, explain the concrete risk, and recommend a resolution or small set of real alternatives with a preferred choice.
Rank findings by impact and end with a direct readiness assessment. Say explicitly when the approach is sound and no consequential gap remains. Do not edit artifacts.
## Findings
### [high|medium|low] <title>
- Artifacts: <exact sections and repository evidence>
- Type: <inconsistency | missing decision | challenged decision>
- Risk: <concrete consequence>
- Recommendation: <preferred resolution and material alternatives>
## Open Questions
<Only decisions needed to resolve findings, or "None.">
## Overall Assessment
<Whether the definition is ready for task planning and why.>
Signals
- GitHub stars
- 30
- Forks
- 1
- Last commit
- Sep 2026
Advanced
- Catalog kind
- skill
- Gateway key
review-approach- Source
- github.com/codagent-ai/agent-skills