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 thorough, highly actionable lint skill with strong sequencing and validation for its destructive consolidate mode. The main weaknesses are length (inline report/output templates that could be factored out) and a handful of checks described procedurally rather than as executable commands.
Suggestions
Move the full consolidation-report markdown template and the dry-run output example into a reference file (e.g. references/consolidate-template.md) and link to it, reducing SKILL.md length and improving progressive_disclosure.
Convert the remaining procedural checks (orphaned pages, broken wikilinks, contradictions) into concrete grep/glob command snippets like the synthesis-gaps section already provides, to lift actionability from 4 to 5.
Tighten the duplicated log-entry format strings (the LINT and LINT_CONSOLIDATE lines) into a single referenceable template to trim redundant tokens.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and almost entirely actionable with no basic-concept padding, but at ~600 lines it includes long inline templates (consolidation report, dry-run output, full output-format block) that could be trimmed or factored out. | 4 / 5 |
Actionability | Most checks give executable specifics (exact grep patterns, thresholds like 'AMBIGUOUS > 15%', 'cohesion < 0.15', real bash for trust-check/trust-record), but a few checks are described as procedural steps rather than copy-paste commands (e.g., contradictions, orphan detection). | 4 / 5 |
Workflow Clarity | Sequenced 'Before You Start' steps, ordered lint checks, and a consolidate mode with explicit validation checkpoints (dry-run preview, user confirmation, git pre-write snapshot, trust-check --strict) plus error-recovery feedback loops for a batch/destructive operation. | 5 / 5 |
Progressive Disclosure | Well-sectioned single file with clearly signaled one-level-deep references to llm-wiki/SKILL.md for config/provenance/lifecycle detail; no bundle files exist, and large inline output templates are the main organization gap. | 4 / 5 |
Total | 17 / 20 Passed |