Content
67%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 well-structured, instruction-only skill whose workflow is clearly sequenced with explicit done-when gates and concrete tool/path guidance. Its main weaknesses are redundancy — sibling-skill routing and the registry-events propose wording each repeated three times — and the absence of a sample output table or script invocation that would make the expected deliverable fully concrete.
Suggestions
State the sibling-skill routing (positioning-mapper, audience-belief-mapper, keyword-research, offer-claims-registry) once — in the Scope guard — and have Reference Materials and Next Best Skill link back to it instead of restating each handoff with new annotations.
State the claims-routing rule ("mark `[needs source]` and route to `memory/events/claims.ndjson` via an authorized `operation: propose` request to `registry-events.py`") once in the Skill Contract, then reference it briefly from Instructions step 4 and Save Results rather than repeating the full wording verbatim three times.
Add a one-row example of the per-competitor teardown table (arc, claimed onlyness sentence, proof pattern, primary framing, source label) and a sample invocation of `scripts/connectors/firecrawl.py` so the expected output shape and commands are copy-paste concrete.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The instruction steps are lean and assume competence, but the same routing content is repeated: sibling-skill handoffs appear in the Scope guard ("reuse positioning-mapper... audience-belief-mapper... keyword-research..."), again in Reference Materials, and a third time in Next Best Skill, while "an authorized `operation: propose` request to `registry-events.py`" appears verbatim three times (Skill Contract Writes, Instructions step 4, Save Results). This matches the 3 anchor ('mostly efficient but includes some unnecessary explanation or could be tightened') rather than 4, where trimming would be only minor. | 3 / 5 |
Actionability | Concrete, executable guidance throughout: named tools ("Scrape with `scripts/connectors/firecrawl.py` and label each row Measured with the source URL"), exact paths ("write `memory/narrative/category-narrative-mapper/YYYY-MM-DD-<topic>.md`"), and a precise labeling convention (Measured / User-provided / Estimated / proxy). Falls short of the 5 anchor because no script invocation examples or a sample teardown-table row/output template are shown — the expected output is described only in prose. | 4 / 5 |
Workflow Clarity | Seven clearly sequenced steps with checkpoints: a robots pre-flight on every scrape, "Done when" exit criteria ("every named competitor has a teardown row... sourced with a Measured URL or marked User-provided/Estimated"), and ask-before-write gates ("ask before writing", "Do not write memory without asking"). Not a 5 because there are no error-recovery feedback loops (e.g., what to do when a competitor has no archived snapshot or a scrape fails), leaving minor validation gaps at the 4 anchor. | 4 / 5 |
Progressive Disclosure | No bundle files exist alongside this SKILL.md, and the body is a well-sectioned overview with one-level-deep references, each link annotated ("[tale-benchmark.md](../../../references/tale-benchmark.md) — TALE framework; this skill feeds the `T` category frame..."). Good structure with mostly clear references, but the same links surface in three separate sections (Scope guard, Reference Materials, Next Best Skill) and the referenced paths (CONNECTORS.md, scripts/connectors/*.py) live outside this bundle and cannot be verified — minor organization gaps at the 4 anchor rather than the clean split of the 5 anchor. | 4 / 5 |
Total | 15 / 20 Passed |