Content
75%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.
A highly actionable, well-structured procedure skill: exact tool invocations, field IDs, scoring table, and output templates make it copy-paste executable with a clear step sequence and edge-case handling. Its main costs are redundancy — blocker semantics and eligibility rules are each restated multiple times, once with conflicting wording — which hurts token efficiency and slightly muddies the workflow.
Suggestions
State the blocker-detection rule exactly once (e.g., in step 3) and delete the two restatements in the Important section; the current dual formulations ('task appears in outwardIssue' vs 'inwardIssue.key matches the current task') can be read as contradictory.
Trim the Important section to only rules not already covered by steps 4–5 (the tie-break preference is a good candidate to keep).
Consider moving the three result-table templates to a small references file to slim the main body if the skill grows.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The blocker-link semantics are explained three times with differing wording (step 3's "Look for links where the task appears in outwardIssue... In practice: check for links where inwardIssue.key matches..." plus the Important section's "The issuelinks field contains both inward and outward links"), and the Important section restates the eligibility and blocker-resolution rules from steps 4–5. The bulk is dense and efficient, so this fits anchor 3 ("some unnecessary explanation or could be tightened") rather than the noticeably padded anchor 2. | 3 / 5 |
Actionability | The body is fully executable: the exact MCP call `mcp__claude_ai_Atlassian__searchJiraIssuesUsingJql` with cloudId, JQL, fields, and maxResults; a t-shirt-size-to-effort mapping table; the ROI formula with null handling; tie-break rules; and copy-paste-ready output templates including the empty-results message — matching anchor 5's "specific examples cover the common cases". | 5 / 5 |
Workflow Clarity | Eight numbered steps form a clear parse → fetch → analyze → filter → score → present sequence with edge-case handling (ask for input when absent, report when no eligible tasks, tie-breaking). It is a read-only skill so the destructive/batch validation cap does not apply, but step 3's two conflicting formulations of inward/outward link semantics leave a minor gap versus anchor 5's fully explicit checkpoints. | 4 / 5 |
Progressive Disclosure | The body is well organized into Preflight / Inputs / Steps / Important with a single-level structure, no nested references, and no bundle files whose placement could be mishandled. It sits at anchor 4 rather than 5 because the ~110-line body keeps everything inline (e.g., result-table templates) where a small reference file could offload detail, and the over-50-line simple-skill exception does not apply. | 4 / 5 |
Total | 16 / 20 Passed |