Content
70%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 orchestration skill with clear phased workflows, explicit verification checkpoints, and a repair loop for every provider path. Its weaknesses are token efficiency — three provider adapters repeat nearly identical phase lists that a shared spec plus a per-host dispatch table would collapse — and progressive disclosure, with all per-provider detail inlined in one long file rather than split into one-level-deep references.
Suggestions
Collapse the Codex, OpenCode, and Grok adapters into a single shared phase spec plus a short per-host dispatch table (which subagent tool, which subagent types, fallback behavior) to remove ~80 lines of near-duplicate text.
Move the per-provider adapter details into references/ files (e.g., adapters/codex.md) and keep a compact routing summary inline, so the main body stays a navigable overview.
Either include or clearly signal the external 'skills/hyperflow/security.md' blocklist dependency so the reference is verifiable rather than dangling.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and free of concept over-explanation, but the Codex, OpenCode, and Grok adapters repeat nearly verbatim phase lists — 'Read the same .hyperflow/ cache files listed above when present', 'Write or update .hyperflow/tasks/<slug>.md', 'Run a separate verification pass for each completed unit', 'Commit each accepted unit separately using conventional commits', 'Never use --no-verify' each appear three times (~80 lines of near-duplicate). This matches anchor 3 (mostly efficient but could be tightened); not 2 because nothing explains concepts Claude already knows and every line is actionable, not 4 because the duplicated adapter sections are a clear trimming opportunity. | 3 / 5 |
Actionability | Guidance is concrete and executable: named cache files ('.hyperflow/profile.md, .hyperflow/architecture.md...'), durable task files ('.hyperflow/tasks/<slug>.md'), hard rules ('Never use --no-verify; never force-push to main or master'), per-host dispatch specifics ('spawn_subagent... subagent_type: general-purpose / explore'), and a copy-paste prompt skeleton. This matches anchor 4 (mostly executable with minor gaps); not 5 because key routing inputs (what triage's 'flow=deep'/'scope=system-wide' looks like, how to detect exposed subagent tools) are referenced but not specified. | 4 / 5 |
Workflow Clarity | Each provider section gives a clearly ordered 5-phase sequence with explicit validation checkpoints: adversarial verification 'for each completed unit before reporting it', quality gates running 'lint, typecheck, build, and relevant tests', and a bounded repair loop ('Retry focused fixes only for verified failures'). Routing exclusions and fallbacks are enumerated. This matches anchor 5 (clear sequence, explicit validation, feedback loop for error recovery); not 4 because checkpoints and the repair loop are explicit rather than merely present. | 5 / 5 |
Progressive Disclosure | No bundle files exist (references/, scripts/, assets/ are all absent), so all content is in a single 165-line SKILL.md. Structure is clear via headers, but the ~100 lines of per-provider adapter detail could live in one-level-deep reference files behind a compact routing table, and the cited 'skills/hyperflow/security.md' path is not part of this bundle and is only mentioned in passing. This matches anchor 3 (some structure, content that could be separate is inline); not 4 because references are not well-signaled and the bulk detail is inlined. | 3 / 5 |
Total | 15 / 20 Passed |