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 serves an important meta-purpose (ensuring skills are invoked) but is severely undermined by extreme verbosity and repetition. The same core message ('always check skills before responding') is restated in at least 5 different ways including a 12-row table of rationalizations. The skill catalog is the most valuable part and is well-structured, but the surrounding content wastes significant token budget on what amounts to motivational hectoring that Claude doesn't need.
Suggestions
Reduce the Red Flags table to 2-3 representative examples or replace with a single concise rule like 'If you're about to act without checking skills, stop and check first.'
Remove the EXTREMELY-IMPORTANT block and DOT graph — consolidate the core rule into a single clear sentence at the top: 'Before any response or action, invoke the Skill tool for any potentially relevant skill.'
Move the platform adaptation section and instruction priority section to a separate reference file, keeping only the Claude Code instructions inline since that's the primary platform.
Add a brief note on what to do when skill invocation fails or when two skills conflict beyond the simple priority ordering.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Extremely verbose and repetitive. The 'Red Flags' table contains 12 rows that all say the same thing ('check for skills first'). The EXTREMELY-IMPORTANT block, the repeated emphasis on '1% chance', and the extensive rationalization table are all redundant ways of saying 'always check skills first'. The DOT graph adds tokens for a simple linear flow. Claude doesn't need to be told 12 different ways not to skip skill checks. | 1 / 3 |
Actionability | The skill catalog table is concrete and actionable — it provides exact skill names and when to use them. The flow diagram and priority ordering give some structure. However, the core instruction ('invoke relevant skills before responding') is simple and doesn't need the elaborate scaffolding. The actual 'how' (use Skill tool with the name value) is clear but buried under excessive motivational content. | 2 / 3 |
Workflow Clarity | The DOT graph provides a clear sequence for the skill-checking workflow, and the priority ordering (process skills first, then implementation) is helpful. However, there are no validation checkpoints — no guidance on what to do if a skill invocation fails, returns unexpected content, or conflicts with another skill. The subagent-stop escape hatch is mentioned but not integrated into the workflow. | 2 / 3 |
Progressive Disclosure | The skill catalog serves as a well-organized reference table pointing to individual skills, which is good progressive disclosure. References to platform-specific tool mapping files (copilot-tools.md, codex-tools.md) are appropriate. However, the main body content is monolithic — the Red Flags table, instruction priority section, and skill types section could be trimmed or moved to a reference file rather than loaded into every conversation. | 2 / 3 |
Total | 7 / 12 Passed |