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.
An exceptionally actionable behavioral skill with a clearly sequenced workflow, explicit validation gates, feedback loops, and a closing checklist. The main cost is token weight from deliberately repeating core rules across the Process, Rationalizations, Red Flags, and Verification sections.
Suggestions
Consolidate the repeated stop rule and "whatever you think" handling: state each once in The Process and have the Rationalizations table and Red Flags reference the step rather than restate it, cutting the biggest source of duplicated tokens.
Move the worked example dialogue (about 30 lines) to a references/example.md file and link it from the Process section, which would simultaneously reduce body size and improve progressive disclosure.
Trim the Verification checklist from 9 items to the 4-5 that aren't already direct restatements of Red Flags entries.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and assumes Claude's competence (no basic-concept padding), but key rules are repeated across three or four sections — the "predict the next three questions" stop rule appears in Steps 1, 6, the rationalizations table, red flags, and the checklist, and "whatever you think" handling appears in Step 5, the table, and red flags. Matches 'mostly efficient but could be tightened'; not a 4 because the repetition is more than minor trimming. | 3 / 5 |
Actionability | Fully executable for an instruction-only skill: copy-paste-ready formats for HYPOTHESIS/CONFIDENCE, the Q/GUESS pair, the six-field restate template, exact scripted responses to each non-yes answer, a concrete save path (docs/intent/[topic].md), and a complete worked example dialogue. Not a 4: no gaps in concrete guidance for the common cases. | 5 / 5 |
Workflow Clarity | Six clearly numbered steps with an explicit confirmation gate ("explicit yes, not 'whatever you think'"), a checkable stop test, an error-recovery feedback loop ("If they correct you, fold the correction in and restate. Loop until an explicit yes"), an escalation floor, and a final verification checklist. Matches the anchor for clear sequence with validation steps, feedback loops, and a checklist. | 5 / 5 |
Progressive Disclosure | No bundle files exist, so all content is inline in a single ~230-line file, but it is well-sectioned with clear headers (When to Use, Process, Example, Red Flags, Verification) and no nested references. Not a 5: it exceeds the short-skill range where a monolithic file scores full marks, and the rationalizations table and example dialogue are borderline candidates for a separate reference file. Not a 3: navigation is easy and nothing is buried. | 4 / 5 |
Total | 17 / 20 Passed |