Content
65%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The body is highly actionable, with executable gh/GraphQL commands and a concrete sub-agent template organized into a clear five-step workflow. Its weaknesses are repetition in the reply/resolve blocks, the absence of a fix-verification checkpoint in a batch editing workflow, and large inline blocks that could be split into reference files.
Suggestions
Factor the repeated agentic-reply message and resolveReviewThread blocks into a single parameterized template to remove triplicate duplication.
Add a verification checkpoint after sub-agent fixes (e.g., run tests / re-read the changed region) before resolving threads on GitHub.
Move the sub-agent prompt template and GraphQL queries into reference files (e.g. references/subagent-prompt.md, references/graphql-queries.md) and link to them from SKILL.md to improve progressive disclosure.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly efficient and actionable, but the GraphQL resolveReviewThread query and the three near-identical agentic-reply message blocks (FIX/IGNORE/NEEDS_MANUAL_REVIEW) are repeated verbatim and could be factored into a shared template. It is not 3 because of this repetition and the long inline sub-agent prompt, and not 1 because it avoids explaining concepts Claude already knows and stays instruction-focused. | 2 / 3 |
Actionability | It provides fully executable `gh pr view`, `gh api`, and `gh api graphql` commands, a concrete copy-paste sub-agent prompt template, and a precise report table format. It is not 2 because the guidance is complete and executable rather than pseudocode or abstract direction. | 3 / 3 |
Workflow Clarity | The five steps are clearly sequenced with conflict-avoidance guidance, but this batch operation (parallel sub-agents editing files) lacks a verification checkpoint confirming fixes are correct or safe before resolving threads. Per the guidelines, missing validation in batch/destructive workflows caps this at 2. It is not 1 because the sequence and per-thread handling are explicit, and not 3 because no validate-fix-then-verify loop is present. | 2 / 3 |
Progressive Disclosure | The skill is a single well-sectioned file with no nested references, but at over 200 lines the large inline sub-agent prompt template and GraphQL query blocks are content that could reasonably be split into reference files. It is not 3 because the simple-skill exception does not apply and substantial inline blocks could be externalized, and not 1 because organization into labeled steps is clear with no deep reference nesting. | 2 / 3 |
Total | 9 / 12 Passed |