Content
50%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 operational core of this skill is strong — concrete, copy-paste-ready Membrane CLI commands with clear connection-state handling and a sensible actions-to-proxy escalation path. Its major defects are a massively padded, largely duplicated 170-line entity list that wastes context, a broken "Step 2" reference, and the absence of any validation guidance for the destructive delete operations the skill exposes.
Suggestions
Delete the ~170-line CodeREADr Overview entity list (or deduplicate and move it to a references/ file) — it is over half the body, contains many repeated and irrelevant entries, and adds no actionable guidance; the popular-actions table already conveys the real domain objects.
Add a validation step before destructive operations (delete-scans, delete-database, delete-user), e.g. list/verify the target resources with the corresponding list action and confirm with the user before running the delete — this would lift the workflow-clarity cap.
Fix the internal navigation: the text says "skip to Step 2" but no Step 2 exists; either label the sections as steps (Step 1 connection, Step 2 search/run actions) or rewrite the references to point at the actual section names.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The ~170-line "CodeREADr Overview" entity list is heavily padded — many entries (Event, Task, Location, Trend, Anomaly, Outlier, Correlation, Regression, Classification, Clustering, Prediction, Recommendation, Sentiment, Emotion, Intent, Context, Insight, Pattern) appear twice, and most (Video, Playlist, Chatter Post, Campaign, Gauge) are irrelevant filler — plus the no-op line "Use action names and parameters as needed." The Membrane CLI sections themselves are efficient, so this sits between the severe (1) and noticeable (2) verbosity anchors; the operational half not being padding places it at 2. | 2 / 5 |
Actionability | Concrete executable commands throughout: "npm install -g @membranehq/cli@latest", "membrane connection ensure \"https://www.codereadr.com/\" --json", "membrane action list --connectionId=CONNECTION_ID --intent \"QUERY\" --limit 10 --json", "membrane action run <actionId> --input '{\"key\": \"value\"}' --json", plus a proxy flag table and a popular-actions table with action keys. Minor gaps — placeholders (CONNECTION_ID, <actionId>) are never filled with a worked example showing how a real action id is obtained and used — keep it below the fully-executable anchor 5. | 4 / 5 |
Workflow Clarity | The install → login → connection ensure → poll state → search actions → run → proxy fallback sequence is clear, with explicit state checkpoints (READY, BUILDING, CLIENT_ACTION_REQUIRED with clientAction fields, CONFIGURATION_ERROR/SETUP_FAILED with error checking). Two flaws hold it at anchor 3: the text says "skip to **Step 2**" twice but no Step 2 is labeled in the document, and the action set includes destructive/batch operations (delete-scans, delete-database, delete-database-value, delete-user) with no validation or verification guidance — the rubric caps workflow clarity at 3 for destructive/batch skills lacking validation. | 3 / 5 |
Progressive Disclosure | No bundle files exist and the body is ~310 lines, well past the under-50-line simple-skill exception. Section headers are well organized, but the 170-line entity list is bulk inline content that belongs in a separate reference file (or should simply be removed), matching the anchor for content that should be separate sitting inline within otherwise-present structure — not the good-placement of anchor 4 nor the header-less monolith of anchor 2. | 3 / 5 |
Total | 12 / 20 Passed |