Content
78%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 navigation-hub skill: concise routing tables, two clearly sequenced default workflows with explicit fallback criteria, and an exemplary progressive-disclosure layout where every reference link resolves to a real, flat, substantive file. The main gaps are minor — some duplication between the two tool tables and no error-recovery guidance for polling failures.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and assumes competence — routing tables, exact endpoint names, and pointers, with zero explanation of concepts Claude already knows. Minor duplication keeps it below 5: the MCP tools appear in both the 'Tool Selection' and 'Quick Reference' tables, `prompt-optimizer.md` is linked five times, and the deprecation notice repeats the frontmatter's wording. | 4 / 5 |
Actionability | Guidance is concrete and executable: named MCP tools with their parameters ('duration_sec, orientation, avatar_id'), exact endpoints ('POST /v1/video_agent/generate', 'GET /v2/videos/{video_id}'), and the auth header ('X-Api-Key: $HEYGEN_API_KEY'). It falls short of 5 because no request-body or response examples appear in the body — a user must open the references to construct a complete call, which is acceptable but leaves minor gaps. | 4 / 5 |
Workflow Clarity | The 'Default Workflow' gives clearly sequenced numbered steps for both the MCP and direct-API paths, with status polling ('poll status and get the download URL') acting as a natural checkpoint, and explicit criteria for when to fall back to the v2 API ('Only use v2/video/generate when user explicitly needs...'). Not 5: there is no error-recovery guidance (what to do on a failed video status, credit exhaustion, or polling timeout), so feedback loops are implicit rather than explicit. | 4 / 5 |
Progressive Disclosure | The body is a clean routing overview: all 22 referenced files exist, are substantive (8,200+ lines total), and are exactly one level deep — cross-links among references point to flat siblings, never burying content in nested chains. Navigation is well-signaled via the 'Quick Reference' task table, per-path workflow pointers, and a categorized 'Reference Files' index. Every piece of detail (visual styles, scripts, captions, webhooks) is appropriately split out of SKILL.md. | 5 / 5 |
Total | 17 / 20 Passed |