Content
85%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.
A highly actionable, well-sequenced workflow with explicit validation checkpoints and copy-paste commands throughout. The main weakness is structural: the dense tasks.md formatting rulebook is inlined into SKILL.md with no progressive disclosure, which both bloats the always-loaded context and buries the rules an implementer needs.
Suggestions
Move the worktrail-compile tasks.md formatting rules (the `files:` continuation syntax, `[e2e]`/`[cleanup]` tags, hot-file ownership bias, per-phase splitting, `review: skip`, `depends:`) into a references/tasks-format.md and keep a 5-10 line summary plus a clearly signaled link in SKILL.md, loading the detail only when the tasks artifact is being written.
Remove the closing guardrail paragraph that re-states the tasks.md rules almost verbatim; point at the same single source of truth instead so the two cannot drift apart.
Trim rationale asides inside the rules block (e.g., "This turns what would be a cross-phase collision into independent per-phase file scopes") to bare instructions — the reasoning can live with the rules in the extracted reference file.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dominated by copy-paste commands and novel worktrail-specific rules Claude could not know ("worktrail-compile recognizes an optional indented `files:` continuation line"), so most tokens earn their place. Minor trimmable material remains: the closing guardrail re-states the requirement-coverage, file-less-task, sizing, and `review: skip` rules almost verbatim, and a few rationale asides ("This turns what would be a cross-phase collision into...") could be tightened — matching anchor 4 rather than the lean anchor 5. | 4 / 5 |
Actionability | Every step carries a copy-paste command ("openspec new change", "openspec status --change \"<name>\" --json", "openspec instructions <artifact-id> --change \"<name>\" --json") with the JSON fields to parse spelled out, plus a concrete formatted example of the `files:` continuation line. This is fully executable guidance for the common cases, matching the anchor-5 example. | 5 / 5 |
Workflow Clarity | Steps 1-5 are clearly sequenced with explicit validation checkpoints: re-run "openspec status --change \"<name>\" --json" after each artifact to check `applyRequires` completion, "Verify each artifact file exists after writing before proceeding", and a pre-final-status checklist re-checking tasks.md against every rule. Error-recovery loops are present (AskUserQuestion on unclear input). This matches anchor 5; the workflow is generative, not destructive or batch, so the validation cap does not apply. | 5 / 5 |
Progressive Disclosure | There are no bundle files at all, and the ~60-line block of worktrail-compile tasks.md formatting rules (`files:`, `[e2e]`/`[cleanup]` tags, per-phase hot-file ownership, `depends:` edges) is exactly the dense reference material that belongs in a separate one-level-deep file like TASKS_FORMAT.md. Section headers exist (Steps, Output, Guidelines, Guardrails), so structure is present but content that should be separate is inline — anchor 3 rather than 2. | 3 / 5 |
Total | 17 / 20 Passed |