Content
77%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The body is a strongly actionable, well-sequenced instruction skill with concrete MCP commands, explicit confirmation gates, and thorough failure/rate-limit handling. Its weaknesses are duplication (devil's-advocate and cache-policy text repeated across sections) and a bundle-integrity defect: the five `agents/*.agent.md` files it routes to are missing from the bundle, breaking progressive disclosure at the exact point where detail is most needed.
Suggestions
Ship or fix the missing `agents/*.agent.md` files — the skill's Step 1 routing depends on them, and none exist in the bundle.
Deduplicate repeated text: state the devil's-advocate escalation conditions once (Step 1) and the label/assignee cache policy once (Step 6) instead of repeating both across sections.
Consolidate repository-resolution guidance, which currently appears in both "Required inputs" and Step 2, into a single location.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Most content is operational knowledge Claude would not already know (MCP tool names, cache policies, rate-limit costs, mutation ordering), but there is real duplication that could be tightened: the devil's-advocate escalation conditions appear nearly verbatim in both the "Agent modes" section and Step 1, and the label/assignee cache policy is stated in "Required inputs" and again in full in Step 6. This matches the anchor "mostly efficient but could be tightened"; it is not a 2 because there is no padding or explanation of concepts Claude already knows. | 3 / 5 |
Actionability | Guidance is fully executable for an instruction-only skill: exact MCP tool names (`mcp_github::issue_write`, `mcp_github::search_issues`, `mcp_github::list_issue_types`), an ordered mutation sequence with payload fields spelled out, concrete cache file paths, template fallback paths under `assets/issue-templates/`, and worked examples of closing keywords (`fixes owner/repo#123`). It is not a 4 because there are no material gaps — the common cases (create, update, sub-issue linking, assignee/label handling, failure recovery) are all specifically covered. | 5 / 5 |
Workflow Clarity | The eight steps are explicitly sequenced with validation checkpoints and feedback loops: duplicate search before drafting (Step 3), draft review and explicit publish confirmation (Step 5), pre-mutation gate confirmation (Step 6), and error-recovery paths for mutation failure (explain, point to `references/mcp-server.md`, retry after user confirms) and rate limits (respect `retry-after`/`x-ratelimit-reset`, stop non-essential lookups). This matches the top anchor; the only adjacent anchor (4) is ruled out by the presence of explicit validation steps and feedback loops throughout. | 5 / 5 |
Progressive Disclosure | The body is well-organized with clearly signaled one-level-deep references (`references/instructions.md`, `references/mcp-server.md`, `references/questions.md`, and the `assets/` templates, all of which exist), but the central routing mechanism points to `agents/bug.agent.md`, `agents/feature.agent.md`, `agents/user-story.agent.md`, `agents/task.agent.md`, and `agents/devils-advocate.agent.md` — and no `agents/` directory exists anywhere in the bundle. The skill's primary execution targets are broken references, which is a larger navigation defect than the "minor organization gaps" of the 4 anchor, though the structure and signaling of the references that do exist keep it above the "references buried / content inlined" problem of the 2 anchor. | 3 / 5 |
Total | 16 / 20 Passed |