Conciseness

The body is lean and every section earns its place: the two-sentence intro conveys what AKF is (a tool Claude would not know), commands are given as bare code blocks with no filler, and evidence examples are one-liners. This matches the anchor 'Lean and efficient; assumes Claude's competence; every token earns its place' — there is no explanation of concepts Claude already knows and no padding to trim.

Actionability

Concrete, copy-paste-ready commands cover the main cases: `akf stamp <file> --agent <agent-name> --evidence "..."`, `akf read <file>`, `akf inspect <file>`, `akf audit <file> --regulation eu_ai_act|hipaa|sox|nist_ai`, labels, and install. This fits 'Mostly executable guidance; concrete code or commands with minor gaps' — it falls short of 5 because expected output of read/inspect is not shown, so Claude cannot verify what a successful stamp or audit looks like.

Workflow Clarity

Sections sequence the workflow by lifecycle stage — check existing metadata 'Before modifying existing files' (`akf read`/`akf inspect`), then stamp 'After creating or modifying files', then audit — which matches 'Clear sequence with most checkpoints present; minor validation gaps'. It is not 5 because there is no explicit verification step after stamping (e.g., re-reading the file to confirm the metadata was embedded), and not 3 because the pre-modification check does act as a checkpoint and each stage's commands are fully specified.

Progressive Disclosure

The skill body is under 50 lines with no bundle files (no references/, scripts/, or assets/ exist), and content is appropriately inline with clear, well-organized section headers. Per the rubric's simple-skill guideline ('For skills under 50 lines with no need for external references, progressive disclosure can score 5 with just well-organized sections'), this earns a 5 — nothing that belongs in a separate file is inlined, and navigation is trivial.

Specificity

The description lists several concrete actions — "stamps every file with trust scores, source provenance, and compliance metadata", "Embeds into 20+ formats (DOCX, PDF, images, code)", and "EU AI Act, SOX, HIPAA auditing" — comparable to the anchor 'Lists several specific actions; minor gaps in coverage'. It falls short of 5 because the actions skew toward marketing framing ('EXIF for AI', 'The AI native file format') rather than a comprehensive enumeration of capabilities, and above 3 because multiple distinct operations are named, not just a domain.

Completeness

The 'what' is clear (stamping files with trust/provenance/compliance metadata, embedding across formats, regulation-specific auditing), but there is no 'when should Claude use it' clause — no trigger guidance such as 'Use when stamping or auditing AI-generated files'. Per the judging guidelines, a missing 'Use when...' clause or equivalent caps completeness at 3, matching the anchor 'Has a clear what but when is missing or only weakly implied'.

Trigger Term Quality

Relevant keywords are present ("trust scores", "source provenance", "compliance", "EU AI Act", "SOX", "HIPAA", "DOCX, PDF"), but natural user phrasings are missing — no 'stamp files', 'check if AI-generated', 'audit provenance', or any 'Use when...' trigger clause. This matches the anchor 'Some relevant keywords but missing common variations or synonyms' rather than 4, which requires good coverage of the natural terms users would actually say.

Distinctiveness Conflict Risk

The niche — provenance/trust metadata and compliance auditing for AI-generated content — is fairly distinct, anchored by specific trigger terms like 'EU AI Act', 'SOX', 'HIPAA auditing', and 'provenance'. This matches 'Mostly distinct; minor overlap risk with closely related skills' (e.g., compliance or document-metadata skills) rather than 5, since 'stamps every file with... metadata' and 'Embeds into 20+ formats (DOCX, PDF, images, code)' are broad enough to bleed into general file-metadata territory.