Content
93%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 exemplary instruction-only skill body: lean, method-specific, and fully actionable through enumerated fields, unit examples, and verified one-level-deep templates. The single minor gap is that the audit checklist is framed for auditing others' reports rather than being explicitly wired in as a pre-delivery self-check on the workflow's own output.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and assumes Claude's competence: no general definitions (what a repost or export is), no padding, and every line is method-specific instruction ("Write `unknown` for any missing value", "The repeated content of a duplicate group counts once as support"). The one-paragraph intro sets the method frame in three lines; nothing reads as over-explanation of known concepts, so it fits the anchor-5 case rather than the minor-trimming case of 4. | 5 / 5 |
Actionability | For an instruction-only skill the guidance is fully concrete: step 1 enumerates the exact protocol fields to write down, step 3 names the inspected-part units ("the post text, the caption, one comment, the profile, image 2, or video 00:00 to 00:45") and points to the record fields in `references/evidence-record.md`, and step 6 points to the claim-table template in `references/report-template.md`. Per the scoring notes, absence of code is not penalized when guidance is this actionable, and the referenced templates plus acceptance examples cover the common cases. | 5 / 5 |
Workflow Clarity | The eight steps are clearly sequenced, step 1 establishes a protocol checkpoint with an explicit change-recording rule, and the audit section supplies a failure-mode checklist (discovery-only, double-counted, misattributed, beyond-sample, empty-search-as-absence). It falls short of 5 only because there is no explicit instruction to run the audit checklist against the researcher's own claim table before delivering — the validation procedure exists but self-application is implied rather than stated as a checkpoint. It is well above 3 since sequence and checkpoints are present, and this is not a destructive/batch operation so no cap applies. | 4 / 5 |
Progressive Disclosure | The ~55-line body is a method overview, with the bulk detail correctly split into three one-level-deep reference files, each verified to exist with substantive content matching its description. Navigation is well signaled: each reference is cited inline at its workflow step and again in a References section that states what each file contains. References cross-mention each other but nest no deeper than one level, matching the anchor-5 structure. | 5 / 5 |
Total | 19 / 20 Passed |