Content
38%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 skill's operative Membrane CLI guidance (roughly the last 150 lines) is genuinely good — executable commands, connection state-machine handling, an actions table, and sensible best practices. However, it is buried under ~850 lines of irrelevant auto-generated entity padding that dominates the file, wrecks token efficiency, and breaks both navigation and workflow coherence.
Suggestions
Delete the entire "CompanyCam Overview" bullet list (lines 21–897): it contains irrelevant items (Greek gods, exoplanets, foundations, legal doctrines), duplicates entries, and terminates mid-item; at most keep the handful of real CompanyCam entities (User, Company, Project, Photo, Tag, etc.).
Fix the workflow numbering: the text says "skip to Step 2" but no Step 2 heading exists; label the sections as Step 1 (connect), Step 2 (search actions), Step 3 (run actions), Step 4 (proxy fallback) so the sequence is followable.
Add validation checkpoints before destructive or batch operations (e.g. confirm the project ID via get-project before delete-project, and verify responses after proxy writes) to raise workflow clarity.
If the full CompanyCam entity/API surface needs documenting, move it to a references/ file (e.g. references/entities.md) and link to it from a short overview, restoring progressive disclosure.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The "CompanyCam Overview" section is roughly 850 lines of bullet points (Zeus, Kepler-800b, foundations, legal terms, NGOs) that are not merely explanations Claude already knows but entirely irrelevant padding, and it even cuts off mid-item ("**Ke"). This squarely matches anchor 1 ("Severely verbose... heavily padded"); no anchor below exists. | 1 / 5 |
Actionability | The Membrane sections give concrete, runnable commands (npm install, membrane login, connection ensure, action list/run with --input JSON, request with flags) plus a Popular actions table, matching anchor 4 ("Mostly executable guidance; concrete code or commands with minor gaps"). It falls short of anchor 5 only because examples use placeholders (CONNECTION_ID, <actionId>) with no single worked end-to-end example and pagination handling is never demonstrated. | 4 / 5 |
Workflow Clarity | A real sequence exists (install → authenticate → connection ensure → poll state → search actions → run/proxy) with state-machine guidance for CLIENT_ACTION_REQUIRED and error states, but it references a "Step 2" that is never labeled, and destructive actions (delete-project, delete-photo) and raw proxy writes have no validation/verification checkpoints. The destructive/batch cap limits this to anchor 3 ("Steps listed but validation gaps"). | 3 / 5 |
Progressive Disclosure | No references/, scripts/, or assets/ directories exist, and the giant entity list is inlined into SKILL.md when it clearly belongs in a separate file (or nowhere) — the textbook anchor 2 case ("content that clearly belongs in separate files is inlined"). It is not a 1 because the operative sections do have usable headers and a table, but it cannot reach 3 while ~850 lines of irrelevant bulk sit inline. | 2 / 5 |
Total | 10 / 20 Passed |