Conciseness

At 1047 lines / ~8,100 words the body is dense and non-generic (no explanations of concepts Claude already knows), but it duplicates itself: routing rules appear twice ("Do not load planning or create-tasks first" in Routing and again in the same section), Gateway credits rules spread across Escalation, Credential Rules, and a dedicated Gateway credits Preference section, placeholder rules appear in three places, and inserted-node payload semantics are stated in Repair Strategy ("an inserted write/create node replaces the payload flowing into the next node") and again verbatim as Workflow Rules 7. Mostly efficient content that clearly could be tightened — the 3 anchor — rather than padded (not 2) or with only minor trim (not 4).

Actionability

Guidance is nearly fully executable: the exact validate command ("node --import tsx node_modules/@n8n/workflow-sdk/dist/cli/index.js validate <filePath>"), complete TypeScript patterns with correct-vs-wrong wiring examples, exact tool-call signatures, and precise placeholder/locator shapes. It stops short of 5 because "{{GROUPING_GUIDANCE_PLACEHOLDER}}" and "{{TOP_LEVEL_ITEM_CEILING_PLACEHOLDER}}" are unsubstituted template tokens — the grouping criteria and ceiling value the Mandatory Process and Completion sections depend on are literally absent.

Workflow Clarity

The 12-step Mandatory Process has explicit validation (step 7 validate, step 8 build), feedback loops (step 10: edit the same file, re-validate, rebuild), a repair budget ("one repair attempt per unique failure signature"), and a Verification Contract with explicit stop conditions — the destructive/batch validation cap does not apply since validation is present. It misses 5 because the mandatory grouping decision checkpoint (step 6 "See [Node Groups](#node-groups) for the criteria") resolves to a placeholder token with no criteria, and the process is scattered across five interleaved sections (Mandatory Process, Repair Strategy, Escalation, Verification Contract, Completion) that restate one another.

Progressive Disclosure

References are clearly signaled by name and all resolve to real bundle files ("load the `error-workflows` reference", "load the `compositional-workflows` reference", "follow `model-selection`", "load the `credential-recipe-research` skill" — each exists in references/), and the 41KB post-build flow is properly externalized. But the body itself is a 1000-line monolith carrying Expression Reference, SDK Patterns Reference, Gateway credits enumeration, and data-table query heuristics that belong in reference files, and reference chains run two levels deep (SKILL.md → post-build-flow.md → trigger-input-data-shapes.md / config-evals.md). Some structure, but content that should be separate is inline — the 3 anchor.

Specificity

The description lists multiple concrete actions — "Write or edit a workspace source file, run workflow-sdk validate via workspace_execute_command, then call build-workflow with filePath" — and covers every work type in its scope (new one-off workflows, existing-workflow edits, verification repairs, workflow-local data tables, evaluations, model choices, credential setup, post-build verification). Coverage is comprehensive rather than minor-gap, so it fits the 5 anchor over the 4 anchor.

Completeness

Both questions are answered explicitly: what ("Write or edit a workspace source file, run workflow-sdk validate ... then call build-workflow with filePath") and when ("Default path for all single-workflow work: new one-off workflows, existing-workflow edits, verification repairs..." plus "Also load for a workflow's evaluations, model choices, credential setup, and post-build verification"), with explicit exclusions on top.

Trigger Term Quality

Good natural keyword coverage: "new one-off workflows", "existing-workflow edits", "verification repairs", "data tables", "model choices", "credential setup", "post-build verification". It falls short of 5 because the phrasing is orchestrator routing language with internal tool names ("workspace_execute_command", "nodes(action=\"execute\")") and lacks synonyms for casual user phrasings.

Distinctiveness Conflict Risk

The description actively de-conflicts against sibling skills: "Do not load planning or create-tasks first", "Load planning only when multiple coordinated workflows ... require a dependency-aware task graph", "Don't use this skill for explicit one-off tasks ... load one-off-operations", and "load data-table-manager first, then this skill". A clear niche with distinct triggers and minimal conflict risk.