Content
75%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 router body: the intent-routing tables, disambiguation rules, and install/update commands are concrete and immediately actionable, and the on-demand skill pointers follow good progressive disclosure. The main cost is token efficiency — the versioning section and the per-workflow details duplicate content that could be trimmed or delegated to the referenced skills.
Suggestions
Trim or relocate the 'Keeping skills current' section's time-sensitive operational detail (the '--skip-skills flag is currently neutered' status and init's network-fallback behavior) — either to the CLI skill or a dedicated notes section — since it competes with routing-relevant context.
Consolidate the 'Workflow details' section with the cheat-sheet table: the Input/Output/Triggers per workflow largely repeat the table row and disambiguation bullets; keep one authoritative routing summary and defer per-workflow specifics to the workflow skills themselves.
Fix the stray fragment in the /pr-to-video details ('≤ (sweet spot 30–90s)') which reads as a typo, and re-check the remaining workflow entries for similar truncations.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The routing tables and disambiguation bullets are dense and information-rich, but the 'Keeping skills current' section carries operational and time-sensitive detail ('The --skip-skills flag is currently neutered (a temporary measure…)', network-fallback behavior of init) that doesn't earn its tokens here, and the per-workflow 'Workflow details' section substantially duplicates the cheat-sheet table. Mostly efficient, but several sections need tightening. | 3 / 5 |
Actionability | Copy-paste-ready commands throughout ('npx skills add heygen-com/hyperframes --skill <workflow-name>', 'npx hyperframes skills check --json', 'HYPERFRAMES_SKIP_SKILLS=1'), concrete spec defaults stated rather than asked (16:9 vs 9:16 for named vertical destinations, language = the user's), explicit ask-limits ('At most two questions'), and specific trigger examples per workflow. | 5 / 5 |
Workflow Clarity | The routing process is clearly sequenced — confirm input → state spec defaults → match via cheat-sheet → disambiguation rules → check the workflow is installed → install and re-read if missing. Ambiguity handling exists ('Genuinely unclear product-vs-topic… ask one question'), but there is no explicit validation/confirmation checkpoint on the chosen route, leaving minor validation gaps. | 4 / 5 |
Progressive Disclosure | Clean router pattern: a capability map pointing to domain skills loaded on demand, an explicit statement that the router 'does not load their internal references', and one-level-deep, well-signaled pointers. No bundle files exist to verify beyond the body. Minor organization gaps from the overlap between the cheat-sheet table and the workflow-details section. | 4 / 5 |
Total | 16 / 20 Passed |