Content
63%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 dense, highly actionable orchestration protocol with a clear sequence, strong validation checkpoints, and well-signaled phase references. Its weaknesses are redundancy between the intro, Authority, and success-criteria sections, and a bundle whose referenced files and scripts are partly missing, which undermines both executability and navigation.
Suggestions
State each obligation once: the under-exploration framing appears in the intro, 'Authority while active', and the success criteria, and the shortened-run obligations (inline show_html_page report plus side-by-side source comparison) are duplicated in Activation and success criteria — consolidate to a single location and cross-reference it.
Restore 'references/fact-ledger.md' (linked as the single definition of the snapshot-and-locator contract) and 'scripts/report-tools.mjs' (invoked by a dozen commands), or move their content into the existing references so no linked path is dead.
Repair the second-level links in 'research-delivery.md' to the missing 'report-gates.md' and 'deliverable-phase.md', or fold their content in, and move the inlined X/social lane rules and probe-conversion workflow from SKILL.md into the phase references to slim the overview.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | There is no known-concept padding and every line is operative, but the under-exploration framing is repeated three times (intro, 'Authority while active', success criteria) and the shortened-run obligations are stated twice (Activation and success criteria), so the body could be meaningfully tightened. Some operational detail inlined in SKILL.md (the X/social lane rules, the probe-conversion workflow) belongs in the phase references. | 3 / 5 |
Actionability | Guidance is highly concrete — exact tool invocations with arguments ('team_create' with 'inline_spec', 'source snapshot --id S<n> --url <url>', 'probe-conversion <raw-file> <converted-file>'), required artifact fields, and error codes ('conversion_altered_number'). It falls short of fully executable because many commands depend on 'scripts/report-tools.mjs', which is absent from the bundle, and on phases defined only in the reference files. | 4 / 5 |
Workflow Clarity | The high-level sequence is clear (activation → scope → team → fan-out → debate → verify → synthesize → deliver → disband) with a 13-item definition-of-done checklist, delivery gates named in order, and a concrete feedback loop (refused fact → replace snapshot → re-add). The detailed step sequences and checkpoints for Phases 0–6 live in the references rather than the body, leaving minor gaps. | 4 / 5 |
Progressive Disclosure | The body itself is well structured with clearly signaled, phase-ordered references ('Read research-collection.md before scoping... research-delivery.md before verification'), but the actual bundle breaks navigation: 'references/fact-ledger.md' is linked (as the single definition of the snapshot/locator contract) yet missing, the 'scripts/' directory containing 'report-tools.mjs' does not exist, and 'research-delivery.md' links to further missing files ('report-gates.md', 'deliverable-phase.md') two levels deep. | 3 / 5 |
Total | 14 / 20 Passed |