Content
57%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 Membrane CLI guidance with a genuinely well-sequenced connection workflow, but it is padded with a useless 'No description' action table and concept explanations Claude already has, and it offers no safety guidance for the destructive actions it advertises. Splitting the action catalog and proxy reference into files would materially cut tokens.
Suggestions
Populate the Popular actions table with real descriptions or move the full action catalog to a references/ file, keeping only the top actions inline in SKILL.md.
Add validation guidance for destructive operations: verify the target first (e.g., run get-user) and confirm with the user before executing delete-user or delete-organization.
Cut known-concept padding (the 'what is Clerk' intro paragraph and 'Use action names and parameters as needed') and fix broken internal references — 'skip to Step 2' points at no section, and '1b' has no matching '1a'.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The bulk is efficient, copy-paste CLI commands, but there are several padded spots: the opening explains what Clerk is ('a user management and authentication platform...'), 'Use action names and parameters as needed' is pure filler, and the 20-row Popular actions table populated entirely with 'No description' adds bulk without information. This fits 'mostly efficient but includes some unnecessary explanation' rather than the noticeably-verbose anchor below. | 3 / 5 |
Actionability | Commands are concrete and executable throughout (install, login, connection ensure, action list/run with --input, the proxy flags table). It is not a 5 because the run examples use '<actionId>' while the Popular actions table only lists keys, leaving the id-to-key mapping unstated, and no worked example with real parameters (e.g., a paginated list-users) is given. | 4 / 5 |
Workflow Clarity | The connection workflow is well sequenced with state checkpoints (poll with --wait, branch on CLIENT_ACTION_REQUIRED, inspect the error field), but it references a 'Step 2' section that does not exist, the '1b' heading has no '1a', and destructive actions (delete-user, delete-organization) are listed with no verification or confirmation guidance — the destructive-operations cap applies, holding this at 3 despite the good connection sequencing. | 3 / 5 |
Progressive Disclosure | A single-file skill (~160 lines) with reasonable section headers but no bundle files at all; the 20-row action catalog and the proxy flag reference are inline content that clearly belongs in separate reference files. This matches 'some structure but content that should be separate is inline' rather than the good-structure anchor, since the doc exceeds the under-50-line simple-skill exception. | 3 / 5 |
Total | 13 / 20 Passed |