Conciseness

The body is dense with n8n-specific, non-obvious knowledge ($binary/$json split, combineByPosition re-attachment, passthroughBinaryImages limits) and never explains generic concepts Claude already knows — matching the score-4 anchor (efficient, minor instances that could be trimmed). It falls short of 5 because the same rules are restated three times: the opening "three rules", the Anti-patterns table, and the Quick Reference Checklist, plus a closing "Remember" line that repeats them a fourth time.

Actionability

Guidance is copy-paste concrete: executable Code-node snippets (getBinaryDataBuffer, Buffer.from(...).toString('base64') with the full binary slot shape), exact field names (responseFormat: "file", combineByPosition, executeOnce: true, includeOtherFields, binaryPropertyOutput), an ASCII wiring diagram for the Merge pattern, and named MCP tools (n8n_executions) for verification. This matches the score-5 anchor — fully executable guidance covering the common cases.

Workflow Clarity

Multi-step processes are explicitly numbered with checkpoints: the inbound staging sequence includes the synchronization-barrier warning and executeOnce: true to prevent N-runs-per-N-files, and the "Verifying binary survived" section provides a genuine feedback loop (produce execution → inspect per-node output → locate the strip point → apply the fix there). Since validation/verification steps are present, the destructive/batch cap at 3 does not apply; this fits the score-5 anchor.

Progressive Disclosure

Structure is good: a Reference files table with "Read when" conditions, inline "See X.md" signals at point of need, one-level-deep references, and appropriate inline-vs-reference splitting — matching the score-4 anchor. It is not 5 because the four referenced files (BINARY_BASICS.md, AGENT_TOOL_BINARY.md, MERGE_FOR_CONTEXT.md, CDN_REQUIREMENT.md) do not exist anywhere in the bundle (no references/ directory), so the well-signaled navigation dead-ends for a reader; it is above 3 because the signaling and split are genuinely well done as written.

Specificity

The description lists multiple concrete capabilities — "Covers the $binary vs $json split, reading/writing binary, keeping binary alive across transforms with Merge, the agent-tool binary boundary, and the CDN/URL requirement for chat surfaces" — which is comprehensive coverage of the skill's scope. It matches the score-5 anchor (multiple specific concrete actions, comprehensive) and is well above score 4, since no meaningful capability area is missing.

Completeness

It explicitly answers both questions: what ("Handle files and binary data in n8n correctly" plus the enumerated coverage areas) and when ("Use when working with files, images, PDFs... or when an AI agent needs a file as tool input or output — and whenever the user mentions..."). This mirrors the score-5 anchor example's structure almost exactly.

Trigger Term Quality

Natural trigger coverage is exhaustive: "files, images, PDFs, attachments, uploads or downloads, base64, vision/multimodal input" plus user-voice phrases like "read the PDF", "attach the file", "send the image", "Merge losing binary", and technical markers ($binary, binaryPropertyName). This matches the score-5 anchor (comprehensive natural terms including synonyms and platform-specific variations).

Distinctiveness Conflict Risk

The niche is unambiguous — binary handling inside n8n specifically — with distinct trigger markers ($binary, binaryPropertyName, Merge losing binary) that no general file-processing skill would share. Conflict risk with other skills is minimal, matching the score-5 anchor; voice is imperative/third person, so no person-penalty applies.