Content
35%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This skill reads as an internal coordination protocol with heavy domain-specific jargon but lacks the concrete, executable guidance needed for effective skill content. The procedure is abstract rather than actionable, with no code examples, command templates, or structured output samples despite defining a detailed output schema. The progressive disclosure structure references many files but the organization is cluttered and the core content defers too much critical detail to unavailable references.
Suggestions
Add a concrete, complete example of the structured output format (JSON/YAML) showing all fields like schema_version, mode, stage_map, etc. with realistic values so Claude knows exactly what to produce.
Replace the abstract procedure steps with specific, executable instructions — e.g., 'Read .harness/tracker/*.yaml to extract stage status' or 'Run `grep -r stage_status .harness/` to find incomplete stages' rather than 'Reconstruct lifecycle state from live repo'.
Consolidate the repeated 'See references/hot-path-folded-context.md' references into a single note at the top or bottom, and organize the References section into a clean list with one-line descriptions of when each reference applies.
Expand the Examples section with at least one full input→output example showing the complete structured output for a realistic reconciliation scenario.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The skill is moderately efficient but contains significant jargon-heavy prose that could be tightened. Terms like 'context-disposition policy', 'blackboard_delta', and 'coding_lens' are used without definition, and several sections repeat the pattern of listing concerns without adding clarity. The repeated 'See references/hot-path-folded-context.md' lines add noise. | 2 / 3 |
Actionability | The skill provides no concrete code, commands, or executable examples. The procedure is described abstractly ('Reconstruct lifecycle state from live repo...') without specific tool invocations, file paths to inspect, or command-line operations. The examples section describes scenarios in prose rather than showing actual inputs/outputs or structured output examples. | 1 / 3 |
Workflow Clarity | There is a numbered procedure with a logical sequence, and validation/failure handling sections exist with explicit gate statuses (pass/fail/blocked). However, the procedure steps are abstract and lack concrete validation checkpoints between steps. Critical detail is deferred to 'references/hot-path-folded-context.md' which is not provided, making the actual workflow incomplete as written. | 2 / 3 |
Progressive Disclosure | The skill references multiple external files (contract.yaml, evals.yaml, hot-path-folded-context.md, stage-arc-boundary-contract.md, etc.) which shows intent for progressive disclosure. However, none of the bundle files are provided, references are numerous and somewhat disorganized (some inline, some in References section, some in other sections), and the repeated 'See references/hot-path-folded-context.md' pattern across multiple sections suggests content organization could be cleaner. | 2 / 3 |
Total | 7 / 12 Passed |