Content
71%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 strongly actionable, mostly lean CLI-oriented skill: nearly every step is an executable command, and the connection lifecycle is handled with explicit state checks and a retry loop. Weaknesses are the dangling "Step 2" reference, the unclear Key-vs-actionId relationship in the run command, and the inlined bulk of the popular-actions table that belongs in a reference file.
Suggestions
Fix the dangling navigation: "skip to **Step 2**" points to a heading that doesn't exist — replace it with the actual section name (e.g. "Searching for actions").
Move the 20-row popular-actions table to a references file (e.g. references/actions.md), keeping a few representative examples inline, and clarify that the table's "Key" is the <actionId> used by `membrane action run`.
Show one complete example of a create action with a real `--input` JSON payload, and trim the filler ("Use action names and parameters as needed.", the bare entity list under Ambivo Overview) to tighten token efficiency.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dominated by executable commands, code blocks, and tight tables, with only minor trimmable padding (the "customer data platform (CDP)" intro, the near-empty Ambivo Overview section, and filler like "Use action names and parameters as needed."), fitting anchor 4 rather than anchor 5's every-token-earns-its-place. | 4 / 5 |
Actionability | Concrete copy-paste commands cover install, login (including headless completion), connection setup, state polling, action search/run, and proxy requests with a flag table, but the popular-actions table lists a "Key" (e.g. list-orders) while the run command uses <actionId> without relating the two, and no example input is shown for a create action — minor gaps matching anchor 4. | 4 / 5 |
Workflow Clarity | A clear sequence (install → login → connection ensure → poll state → search → run) with a real feedback loop for CLIENT_ACTION_REQUIRED (user completes action, then poll again for READY) and explicit error-state handling; however "skip to **Step 2**" references a heading that doesn't exist in the document, and no validation of action results is described — anchor 4, not 5. | 4 / 5 |
Progressive Disclosure | This is a single ~160-line SKILL.md with no bundle files and well-organized sections, but the 20-row popular-actions table and the full CLI state-machine detail are bulk content inlined that would fit a separate reference file, and the skill is too large for the under-50-lines simple-skill exception — matching anchor 3 (some structure, content that should be separate is inline). | 3 / 5 |
Total | 15 / 20 Passed |