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.
A strong, highly actionable skill body with copy-paste YAML, exact MCP call syntax, and a well-validated workflow including error-recovery loops. Its weaknesses are systematic rule duplication across Instructions, examples, Performance Notes, and Troubleshooting (which inflates the body to ~600 lines), and an orphaned bundle file (references/v0-pipeline-schema.md) that the body never links to.
Suggestions
State each validation rule once — consolidate the failureStrategies requirement, matrix-placement rule, and 'check native-steps.md before Run steps' into a single Validation/Troubleshooting section instead of repeating them across Instructions, stage sections, Performance Notes, and Troubleshooting.
Add a pointer to references/v0-pipeline-schema.md (e.g., in the Pipeline Structure or Common Step Types section) — the 453-line schema reference currently exists in the bundle but is never referenced from SKILL.md, making it undiscoverable.
Trim or merge the two complete examples (Complete CI Example and Complete CD Example), which largely duplicate the stage-type and step-type snippets above them, or move one to a reference file.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly efficient, Harness-specific content with no filler explanations of concepts Claude already knows, but it repeats the same rules multiple times: the failureStrategies requirement appears in Instructions ("Every CI and CD stage must include a failureStrategies array"), again in the Matrix/Troubleshooting sections, and the matrix-placement rule appears twice ("strategy must be at the stage level, as a sibling of spec" and again in Troubleshooting); "Always check references/native-steps.md before using a Run step" is stated three times (Instructions step 3, Common Step Types, Performance Notes). The two complete examples (CI ~70 lines, CD ~100 lines) largely duplicate the stage and step snippets above them. This lands on the score-3 anchor ('mostly efficient but ... could be tightened') rather than 4, since the duplication is systematic rather than a couple of trimmable spots. | 3 / 5 |
Actionability | Everything is copy-paste ready: complete valid YAML blocks for every stage and step type, exact MCP invocations with parameter blocks ("Call MCP tool: harness_create / resource_type: 'pipeline' / body: { yamlPipeline: ... }"), exact identifier regexes ("^[a-zA-Z_][0-9a-zA-Z_]{0,127}$"), and verbatim API error messages mapped to fixes. This fully matches the score-5 anchor (copy-paste ready code/commands covering the common cases). | 5 / 5 |
Workflow Clarity | A clear 5-step sequence (analyze codebase → clarify requirements → select native steps → generate YAML → create via MCP) with explicit validation checkpoints: "First verify the project exists" via harness_list before creating, "To verify it was created" via harness_get afterward, and a Troubleshooting section that maps real API errors to concrete fixes (feedback loops for error recovery). This matches the score-5 anchor (clear sequence, explicit validation, feedback loops) — the requirement-gathering step even forbids guessing placeholder values. | 5 / 5 |
Progressive Disclosure | The body is a well-signaled overview pointing one level deep to references/codebase-analysis.md ("Use the detection tables and decision tree in references/codebase-analysis.md") and references/native-steps.md ("Consult references/native-steps.md for the full mapping"), with the bulk 300+ step catalog correctly split out. However, the third bundle file, references/v0-pipeline-schema.md (453 lines), is never referenced anywhere in SKILL.md — it is orphaned and undiscoverable. That is a real navigation gap beyond the score-5 anchor ('easy navigation'), but the remaining structure is stronger than score 3 ('references present but not clearly signaled'), so score 4 fits best. | 4 / 5 |
Total | 17 / 20 Passed |