Content
82%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 highly actionable, well-organized single-file skill: exact endpoints, parameters, and copy-paste payloads cover the common cases, and the tool surface plus runner_usage rules are unambiguous. Minor gaps: developer-facing maintenance content is inlined, and error conditions are documented without recovery guidance.
Suggestions
Move the 'Implementation & maintenance (for Space developers)' section (internal Go paths, AGENTS.md alignment note) into a separate reference file or drop it — it targets maintainers, not the agent invoking the skill.
Add brief recovery guidance for the documented error states, e.g. 'On 401, verify the PAT is a lbpat_-prefixed Space token; on HTTP 400 from REST, confirm runner_usage is exactly "agent" or "event".'
Consider splitting the detailed runner_usage REST fallback rules (endpoint paths, type_filter/component_filter, webpage URLs) into a one-level reference file, keeping a short summary plus the MCP payload inline.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and assumes Claude's competence (no MCP/Bearer/pagination primers, efficient tool-surface table), but the 'Implementation & maintenance (for Space developers)' section with internal Go file paths serves maintainers rather than the invoking agent and could be trimmed or moved. | 4 / 5 |
Actionability | Everything is executable: exact URLs ('https://space.langbot.app/mcp', self-hosted variant), the exact auth header ('Authorization: Bearer <your-pat>'), copy-paste client-config JSON, a complete tool-surface table with parameter names and accepted values ('chat, embedding, rerank, tts, or asr'), and a ready-to-use search payload. | 5 / 5 |
Workflow Clarity | 'How to use' gives a clear 4-step sequence and error semantics are documented ('Requests without a valid PAT get 401 Unauthorized', 'Invalid REST values return HTTP 400') plus decision rules (select_models availability ordering, runner_usage interpretation), but there is no explicit recovery/feedback loop; the read-only scope keeps risk low so no cap applies. | 4 / 5 |
Progressive Disclosure | Sections are well-organized with clear headers and no buried references, but at ~110 lines with mixed audiences (agent usage vs. Space-developer maintenance) and detailed runner_usage REST-fallback rules inlined, some content could be split into a one-level-deep reference file; the under-50-line simple-skill exception does not apply. | 4 / 5 |
Total | 17 / 20 Passed |