Content
63%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 delivers a well-sequenced, highly actionable reflection workflow with concrete formats, a bias-cue table, and error handling. Its main weaknesses are maintainer-metadata padding, the absence of a worked output example, and a bundle that ships none of the six files the body references — every reference link and script path is dead.
Suggestions
Ship or remove the six referenced bundle files: references/cognitive_bias_canon.md, references/honest_output_discipline.md, references/conversation_reflection_practice.md, and the three scripts/ files in the Tooling table are all absent from the bundle, so every link and script path is dead.
Delete the portability blockquote and the version/source-spec footer — maintainer metadata that does not guide execution and adds token cost.
Add a short worked example (2-3 sentences) of the required flowing-prose output so the tone and format rules have a concrete anchor.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The core framework sections (bias table, recommendation formats, error handling) are tight and skill-specific, but there is real padding: the "This matters because:" rationale bullets under Stop Directive explain what Claude can infer, and the portability blockquote ("Most portable in the v2 collection") plus the version/source-spec footer are maintainer metadata that earns no tokens during execution. Not 4 because these unnecessary sections go beyond minor trimming; not 2 because the bulk is non-redundant, skill-specific guidance rather than explanation of known concepts. | 3 / 5 |
Actionability | Concrete, executable guidance throughout for an instruction-only skill: the exact Q1 wording ("What specifically should I reassess? Pick one..."), the bias recognition-cue table mapping each bias to conversation patterns, exact closing formats ("Continue. {specific reasoning for why}."), and a branch-per-situation error handling table. Not 5 because there is no worked example of the required flowing-prose output to anchor tone, and the Tooling table lists scripts by role only with no usage or commands. | 4 / 5 |
Workflow Clarity | A clear sequence is present: trigger detection (explicit vs implicit, with "don't auto-invoke — ask the user"), halt directive, conditional clarifier with an explicit stop condition ("One question max"), full-conversation re-read, 5-dimension analysis, and a mandatory closing recommendation, plus error-handling branches. Not 5 because there is no validation checkpoint confirming the output satisfies the tone/format rules — the validator script that would provide this feedback loop is described but never wired into the workflow. | 4 / 5 |
Progressive Disclosure | References are clearly signaled and one level deep ("See references/cognitive_bias_canon.md for the full canon"), but no references/, scripts/, or assets/ directories exist in the bundle — all three reference links (cognitive_bias_canon.md, honest_output_discipline.md, conversation_reflection_practice.md) and all three scripts in the Tooling table are dangling paths that resolve to nothing. Not 4 because every advertised bundle file being missing is more than a minor organization gap; not 2 because the inline body is well sectioned and self-sufficient for the core task without the missing files. | 3 / 5 |
Total | 14 / 20 Passed |