Content
81%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 highly actionable, well-validated workflow with strong error recovery and routing, held back mainly by significant redundancy — several rules are stated three to four times across sections — and by inlining detail that could be split into reference files. Tightening repetition alone would substantially reduce token cost without losing any guidance.
Suggestions
Deduplicate the repeated rules: state the 'single fetch_versions.sh call, never loop anypoint-cli-v4', 'never redirect mvn output', and 'absolute script path' rules once (in Rules or the step where they apply) instead of 3-4 times across the Bundled Scripts table, Version Discovery, Step 9, and Rules.
Move the Step 1 project-disambiguation case matrix and/or the Troubleshooting section into a one-level-deep reference file (e.g. references/troubleshooting.md) and keep a short summary in SKILL.md, reducing the ~400-line body toward a lean overview.
Trim rationale prose (e.g. the repeated latency justification for the script and editorializing like 'the single biggest lever for making this skill feel fast') — the instruction is clear without the motivation restated in multiple places.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is operational rather than educational, but the same instructions are repeated many times: the "one script call, never loop anypoint-cli-v4" rule appears in the Bundled Scripts table, Version Discovery intro, and twice in Rules; "never redirect mvn output" appears in both Step 9 and Rules; the absolute-path warning appears twice. The ~400-line body could be tightened considerably, matching the anchor "mostly efficient but includes some unnecessary explanation or could be tightened". | 3 / 5 |
Actionability | Fully executable throughout: exact awk one-liner for pom.xml edits, exact `cd <projectDir> && mvn clean package -DskipTests` command, concrete `fetch_versions.sh` invocation with argument shapes and full JSON output schema, and literal question/response templates. Copy-paste ready with specific examples covering display, check, and change paths. | 5 / 5 |
Workflow Clarity | Clear routing table maps user intent to sequenced step chains, STOP markers define checkpoints, and validation/feedback loops are explicit for the risky operations: pom.xml content backed up in memory and restored on Exchange fetch error or BUILD FAILURE, `BUILD SUCCESS` gating completion, and per-row `source` (list/describe-fallback/error) handling with distinct recovery paths. | 5 / 5 |
Progressive Disclosure | Structure is good: a single bundled script in scripts/ is documented with usage, output shape, and absolute-path invocation guidance, and the Execution Paths table provides navigation. However the SKILL.md itself is ~400 lines with the Step 1 disambiguation matrix and Troubleshooting sections inlined — content that could live in one-level-deep reference files to keep the overview lean, so it falls short of the clear-overview anchor. | 4 / 5 |
Total | 17 / 20 Passed |