fix(skills/copilot-review-loop): report the effort level a review states

[sourcehawk]

Description

Fixes #95

The skill said that nothing in a Copilot review records its effort level, and told the agent not to report that a review was balanced. That statement is false: a review in the current format states its level in a labelled line of its body. This PR tells the agent to report the level from that line, and to report no level when the line is absent.

Changes

• ## Review effort level is a precondition: the agent reads the line **Review effort:** <level> in the review body, reports that level, and names the line as its source.
• Same section: if the line is absent, the agent says that the review does not state its level. A missing line does not mean lite, and a mention of effort in the prose of the body is not the line.
• Step 4 names the review body as the place where a review states its level.

Challenges

Not every review carries the line. Copilot reviews on six pull requests of this repository show two formats. Each review in the current format (#78, #90, #91, #94) carries the line, with or without an emoji before it, and not always as the last line. Reviews in the older format (#28, #32) carry no such line, and one body on #32 mentions the effort level in its prose summary. The rule for the absent line and the rule for a prose mention come from those reviews.

Testing

Each run gave a new agent the skill file and a review body, and asked for the triage report to the user.

On the unedited skill, six runs reported the level from the body: three on a lite review and three on a balanced review. So the failure in the issue did not occur in these runs. But two of the three runs on a balanced review told the user that the skill was wrong, and did not obey it.

With the new text, five runs passed. Two runs on a balanced review reported Balanced and quoted the line, with no argument against the skill. Two runs on a body with no line, and with a summary that reads as thorough, said that the review does not state its level and guessed none. One run on a lite review reported Lite with the line as its source.

No script changed, so the script tests did not run. The JSON files in .claude-plugin/ parse.

🤖 Generated with Claude Code

[Copilot]

🟢 Approval recommended

The focused guidance matches the documented review formats and handles missing labels safely.

0 open findings

What changed in this PR

Fixes #95 by teaching the Copilot review loop to report only explicitly stated review effort levels.

Changes:

• Parses the labelled effort line from review bodies.
• Avoids inferring a level when the line is absent.
• Clarifies that Step 4 reads effort metadata from the body.

File	Description
skills/​copilot-review-loop/​SKILL.md	Documents reliable effort-level reporting during review triage.

🧠 Review effort: Balanced

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.