Content
67%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 delivers strong executable guidance: concrete CLI commands, a state-aware connection workflow with polling and error recovery, and clear best practices. Its main weaknesses are padded introductory material and structural sloppiness — a "skip to Step 2" reference with no Step 2 and a "1b" heading with no "1a" — that make the flow slightly harder to follow than it should be.
Suggestions
Fix the broken step references: define the numbered steps explicitly (e.g., label the connection setup as Step 1 and add a real "Step 2: Discover and run actions" section), or remove the "skip to Step 2" mentions entirely.
Trim the padded intro (who uses BundleIQ and why) and the filler line "Use action names and parameters as needed." to tighten token efficiency.
Consider moving the proxy options table and clientAction schema details into a reference file to keep SKILL.md as a lean overview.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly efficient and command-driven, but includes unnecessary padding: the intro paragraph about "researchers, analysts, and knowledge workers to build a centralized repository of insights and improve decision-making", the filler line "Use action names and parameters as needed.", and the rambling Agent Types sentence. This fits anchor 3 ('mostly efficient but includes some unnecessary explanation or could be tightened') rather than 4, where such sections would be minor and trimmable rather than several. | 3 / 5 |
Actionability | Nearly all guidance is concrete and executable — `npm install -g @membranehq/cli@latest`, `membrane connection ensure` with polling, `action list/run` with `--input` JSON, and a proxy flags table — copy-paste ready for the common cases. Minor gaps (the dangling "skip to Step 2" reference) keep it at anchor 4 rather than 5. | 4 / 5 |
Workflow Clarity | The setup sequence (install → login → connect → wait for ready → discover actions → run) is clear, and connection-state handling provides explicit checkpoints and a re-poll feedback loop, including error recovery for CONFIGURATION_ERROR/SETUP_FAILED. Not anchor 5 because the sequence is marred by a broken cross-reference ("skip to Step 2" with no Step 2 defined, and a "1b" section with no "1a"); not anchor 3 because checkpoints are explicit rather than implicit. | 4 / 5 |
Progressive Disclosure | The single-file skill has well-organized sections (Install, Authentication, Connecting, Searching, Running, Proxy, Best practices) and no nested references, and no bundle files exist to verify. It is above the <50-line simple-skill exception, and content like the proxy options table and clientAction schema could arguably live in a reference file, matching anchor 4 ('good structure; most content is appropriately placed; minor organization gaps') rather than 5. | 4 / 5 |
Total | 15 / 20 Passed |