Have an agent produce a plan before implementing, and review it as the cheapest intervention point in the work. Use before any non-trivial agent task.
Scanned 9/5/2026
Install to Claude Code
npx -y skills add Amey-Thakur/AI-SKILLS --skill agent-plan-review --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Agent Plan Review?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/amey-thakur-agent-plan-review)More formats (shields.io, HTML) on the badges page.
---
name: agent-plan-review
description: Have an agent produce a plan before implementing, and review it as the cheapest intervention point in the work. Use before any non-trivial agent task.
---
# Agent plan review
Reviewing a plan costs a minute; reviewing a wrong implementation costs
an hour and a revert. Asking for the plan first is the highest-value
habit in agent-assisted work.
## Method
1. **Ask for the plan before any edit.** Files to change, the approach,
and the order, so the direction can be corrected before effort is
spent (see agent-plan-execute-replan).
2. **Check it against your intent.** Agents frequently solve a nearby
problem correctly, and the plan is where that becomes visible.
3. **Look for what is missing.** Tests, migrations, error handling, and
documentation are the usual omissions.
4. **Challenge unnecessary scope.** Plans that touch more than needed
produce reviews that are hard to reason about (see
pull-request-size).
5. **Confirm the verification step.** How the agent will know it worked,
since a plan without a check ends in assumed success.
6. **Approve explicitly and let it run.** Once the plan is right,
interrupting each step wastes the benefit of delegation.
7. **Compare the result to the plan.** Divergence is worth
understanding, since it is either discovery or drift.
## Boundaries
A good plan does not guarantee good implementation and must still be
reviewed. Plan review adds a turn, which is not worth it for trivial
changes. Agents can produce plausible plans for approaches that will not
work, which needs your judgement.
Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!