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 dense, highly actionable pre-research playbook with genuine validation gates, executable commands, and a well-structured one-level-deep reference bundle. Its main weaknesses are repetition of the artifact-resolution rule across three sections, awkward accreted section numbering, and a referenced THIRD_PARTY_NOTICES.md that is missing from the bundle.
Suggestions
State the artifact-resolution rule once (in §4.55) and reference it from Non-Negotiables and Finish Criteria with a one-line pointer instead of restating it nearly verbatim three times.
Renumber the Standard Flow into a clean integer sequence and fold the decimal insertions (§3.5, §4.25, §4.5, §4.55, §4.6) into properly ordered steps or a consolidated 'Gates' checklist so the primary workflow is scannable.
Add the missing THIRD_PARTY_NOTICES.md to the bundle (and fix its path in references/query-design.md, which points at `../THIRD_PARTY_NOTICES.md`), or drop the packaging instruction; also briefly note how `$DEEPLINE_API_BASE_URL`/`$DEEPLINE_API_KEY` are expected to be set.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly operational and avoids explaining known concepts, but it is padded by repetition: the artifact-resolution rule appears nearly verbatim in Non-Negotiables ('Resolve every materializable dataset to its exact artifact'), again as the full §4.55 section, and again in Finish Criteria ('Every materializable dataset must be resolved to an exact artifact... not named from memory (§4.55)'), and the decimal section numbering (3.5, 4.25, 4.5, 4.55, 4.6) signals accreted insertions rather than a tightened document. | 3 / 5 |
Actionability | Concrete, executable commands throughout: the `deepline search "topic -> recent web and news sources" --type tools --categories research` examples, the curl calls to `/api/v2/pre-research/plan`, the runnable SKILL_ROOT resolution loop, and exact search query shapes. Minor gaps keep it below 5: `$OBJECTIVE`, `$DEEPLINE_API_KEY`, and `$DEEPLINE_API_BASE_URL` are used without setup guidance, and the report template is a fill-in skeleton. | 4 / 5 |
Workflow Clarity | The Standard Flow is a clear sequence with explicit validation checkpoints and feedback loops (Coverage Gate §4.5, Artifact Resolution Gate §4.55 with 'A cell filled from memory instead of a live result is not verified. Search for it.', Approval Gate before paid runs, Finish Criteria), matching anchor 5's substance. The mid-sequence decimal insertions (§3.5, §4.25, §4.55) make the primary sequence slightly harder to follow, placing it between anchors 4 and 5 but closer to 4. | 4 / 5 |
Progressive Disclosure | References are well-signaled, condition-scoped, and one level deep ('Read `references/source-map.md`', 'For GTM use cases, also read `references/last30days-gtm-corpus.md`; it summarizes...'), and all four reference files plus all three scripts exist in the bundle. Not 5 because `THIRD_PARTY_NOTICES.md` is referenced ('Keep `THIRD_PARTY_NOTICES.md` with this skill when packaging') but absent from the bundle, and substantial inline material (the full route-preference table, the complete report template) could partly live in references. | 4 / 5 |
Total | 15 / 20 Passed |