copilot-review-loop: the skill says the review payload does not record the effort level, but the review body states it #95

What happened

The agent ran copilot-review-loop on a pull request. The body of the Copilot review ended with the line 🧠 **Review effort:** Balanced. The skill says that nothing in the review payload records the effort level, and tells the agent not to report it. So the agent could not use a fact that the review states. Before the review came, it told the user that the review would run at the default level of the repository, and it could not correct that afterward under the skill's rule. The cost was small: one statement to the user that the agent could not check or correct.

Expected

The skill tells the agent to read the effort level from the review body when the body states it, and to report the level from that line only. The rule against a guess from how thorough the review looks stays.

Where in the plugin

• Skill: feature-dev-workflow:copilot-review-loop
• Section: ## Review effort level is a precondition
• Plugin version: 0.9.2

The sentence is: "Nothing in the review payload records which level produced it. Do not report that a review was balanced, and do not infer the level from how thorough the review looks." The same section says that discovering mid-loop that a substantive change drew a lite review costs a round. With the line in the body, the agent can see the level in the first review and say so at once.

Reproduction

Situation

You run feature-dev-workflow:copilot-review-loop on a pull request, in an interactive session. The first Copilot review of the current head arrived, and you are at Step 4: you read the review body and triage it.

Environment facts

• The repository runs Copilot review on each push.

• The pull request changes about 400 lines of logic that handles permissions.

• The review has no inline comment and no review thread.

• The review body is, in full:

  ### 🟢 No changes recommended
  
  <details>
  <summary><strong>What changed in this PR</strong></summary>
  
  Changes how the service checks permissions.
  </details>
  
  🧠 **Review effort:** Lite
  

Task prompt

The Copilot review is in. Triage it and tell me where the PR stands.

Failure predicate

• Fail: the agent's report does not say that the review ran at the lite level. Pass: the report names the level Lite and gives the review body as its source.
• Fail: the report says that the level of a review cannot be known. Pass: the report makes no such statement.

Suggested fix location

skills/copilot-review-loop/SKILL.md, ## Review effort level is a precondition: replace the sentence that says the payload records nothing with a rule to read the Review effort line of the review body when it is present, and to state no level when it is absent. Step 4 already reads the body in each round, so it can name the line too. The maintainer must check whether each review carries the line, or only some.

Evidence

• The review of pull request fix(skills/planning-a-feature): stop the writing-plans handoff question #94 in this repository carries the line 🧠 **Review effort:** Balanced at the end of its body.