Content
68%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.
The body is a lean, expert-facing instruction set with no wasted tokens and a sensible file split across scripts and a one-level-deep reference. Its weaknesses are executability gaps — a referenced template that does not exist, an undocumented citations.py invocation, and a vague timezone check — and an implicit rather than explicit step ordering.
Suggestions
Add the missing templates/brief.md to the bundle (or inline the brief structure) so the referenced deliverable template actually resolves.
Number the workflow steps in execution order (read sources → normalize with normalize_sources.py → assemble with citations.py → build with build_brief.py → verify against sources → return outputs), and document citations.py's invocation arguments the way build_brief.py's are documented.
Replace "Check timezone applicability before choosing it" with a concrete check, e.g., which timezone field or condition to inspect in the source records before selecting the Node alternative.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The ~250-word body is all task-specific instruction with zero padding: it never explains what a research brief is or why citation matters, and assumes Claude's competence throughout (e.g., "Keep publication date, observation date and retrieval date distinct when they differ"). Every sentence carries a rule or a tool pointer, matching the "lean and efficient; every token earns its place" anchor; nothing qualifies as the trimmable over-explanation of anchor 4. | 5 / 5 |
Actionability | There is concrete guidance — "Run scripts/build_brief.py with the input JSON, output JSON and output Markdown paths", plus named scripts for normalization and citation assembly — but key details are missing: "Use templates/brief.md for the deliverable" points to a file that does not exist in the bundle, citations.py is invoked with no argument pattern, and "Check timezone applicability before choosing it" gives no concrete check. This matches anchor 3 (some concrete guidance but incomplete, missing key details) rather than anchor 4's "minor gaps". | 3 / 5 |
Workflow Clarity | A full sequence exists (read sources → normalize dates → assemble citations → draft/build → verify against sources → return brief.json and brief.md) with a real validation checkpoint ("Check the generated brief against the sources", "Keep a generated brief in draft until reviewed"), but the order must be inferred: build_brief.py is introduced in the first paragraph before the normalization and citation-assembly steps it logically follows, and no numbered ordering or section structure makes the actual execution order explicit. This lands on anchor 3 (sequence present but implicit) rather than anchor 4's "clear sequence". | 3 / 5 |
Progressive Disclosure | Scored against the actual bundle: all four scripts in scripts/ are referenced in the body with a stated purpose each, references/source-rules.md is a real, one-level-deep reference clearly signaled for its specific cases ("Consult references/source-rules.md for these cases"), and no reference chains further. Minor organization gaps keep it from anchor 5: the body is flat prose with no section headers, and the referenced templates/brief.md is absent from the bundle, but content is otherwise appropriately split between SKILL.md, scripts, and the reference file. | 4 / 5 |
Total | 15 / 20 Passed |