Content
86%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 efficient, highly actionable SKILL.md body: executable commands for every common path, a sequenced workflow with an explicit validation step and dry-run checkpoint, and clean progressive disclosure to real bundle files that are all one level deep. The only gaps are mild purpose-restatement across three sections and an implicit (rather than explicit) fix-and-retry loop when validation fails.
Suggestions
Add one sentence after workflow step 5 describing the recovery loop, e.g., "If validation fails, fix the ticket (schema bounds, id/folder mismatch) and re-run export + validate" — this would close the workflow-clarity gap.
Merge the purpose statements currently repeated across "Overview", "When to Use", and "Position in Split Workflow" into a single section to trim duplicated tokens.
Drop the reference-file re-listing in workflow step 2 (keep the single "Resources" section) and reference each file only where it is actually needed.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean — no explanations of concepts Claude already knows, and sections like "Export Rules" and "Guardrails" are terse lists where every line is task-specific. Minor trimming is possible: "Overview", "When to Use", and "Position in Split Workflow" each restate the skill's purpose, and the four references are listed in "Workflow" step 2 and again in "Resources", so a few tokens are duplicated. | 4 / 5 |
Actionability | Three complete, copy-paste-ready bash commands cover the common cases (daily auto-detection, export from a ticket, interface-only and interface+schema validation), each with concrete flags like `--ohlcv`, `--export-strategies-dir`, and `--pipeline-root`. The workflow steps name exact script paths, matching the fully-executable anchor. | 5 / 5 |
Workflow Clarity | The 6-step workflow has an explicit validation step ("Validate interface and Phase I constraints with... validate_candidate.py"), a dry-run-first handoff ("run dry-run first"), and a Guardrails checklist with rejection criteria ("Reject candidates that violate schema bounds", "folder name and `id` mismatch"). It falls short of anchor 5 only because the fix-and-retry feedback loop (what to do when validation fails) is left implicit. | 4 / 5 |
Progressive Disclosure | The body is a well-organized overview with a "Resources" section that clearly signals each one-level-deep reference with a one-line description, and every referenced path (`references/pipeline_if_v1.md`, `signal_mapping.md`, `research_ticket_schema.md`, `ideation_loop.md`, and the three scripts) exists in the bundle. No content that belongs in a reference file is inlined, and references are not nested. | 5 / 5 |
Total | 18 / 20 Passed |