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 CLI workflow half of the body is strong — executable commands, a clear auth/connection path, and a well-documented proxy fallback with a flag table. The body is dragged down by a heavily padded 145-line entity catalog (with duplicates and irrelevant entries) that bloats token cost, plus a broken "Step 2" reference that confuses the workflow sequence.
Suggestions
Cut the Bizimply Overview entity list down to the handful of real, commonly-used resources (Locations, Users, Attendance, Absences, Schedules, Payroll Runs) or move the full catalog to a references/resources.md file, removing the duplicated Users/Roles entries and off-domain items like Equity Grants, Journals, and Garnishments.
Fix the two references to a nonexistent "Step 2" (lines 210 and 224) by renaming sections to a consistent Step 1/Step 2 sequence, and standardize on one command form (`membrane` vs `npx @membranehq/cli@latest`).
Add validation guidance for destructive or batch operations (e.g. verify parameters against the action's inputSchema, dry-run or check results via `--json` output before wide-reaching writes) to complete the feedback loop the connection-state section already establishes.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The "Bizimply Overview" is ~145 lines of bare entity names ("Stock Options", "Garnishments", "Fixed Assets", "Journals", "Wellness Programs"...), with duplicated entries (Users and Roles appear both nested under Locations and as top-level items) and zero explanatory value — plus the filler line "Use action names and parameters as needed.". This matches anchor 2 ("Noticeably verbose; several unnecessary explanations or padded sections") — the padding is a major section, though the CLI/auth/proxy sections themselves are reasonably tight, keeping it above 1. | 2 / 5 |
Actionability | Commands are concrete and executable: `npm install -g @membranehq/cli@latest`, `membrane login complete <code>`, `membrane connection ensure "https://www.bizimply.com/" --json`, `membrane action run <actionId> --connectionId=... --input '{...}' --json`, plus a full proxy flag table. Minor gaps: `membrane login --tenant --clientName=<agentType>` leaves `<agentType>`/`--tenant` under-specified, and commands inconsistently switch between `membrane` and `npx @membranehq/cli@latest`. Anchor 4 ("Mostly executable guidance ... minor gaps") fits; 5 is blocked by the unresolved placeholders and command inconsistency. | 4 / 5 |
Workflow Clarity | A real sequence exists (install → login → connection ensure → poll to READY → search actions → run/proxy) with a state-machine feedback loop for connections (BUILDING/READY/CLIENT_ACTION_REQUIRED/CONFIGURATION_ERROR), but it is undermined by two references to a nonexistent "**Step 2**" (only a "1b" heading exists), inconsistent command forms, and no validation guidance for destructive or batch action runs. Anchor 3 ("Steps listed but validation gaps; sequence present but checkpoints missing or implicit") fits: better than 2 because the connection-state checkpoints are explicit, short of 4 because the dangling Step-2 references break the sequence and risky operations lack validation. | 3 / 5 |
Progressive Disclosure | No bundle files (references/, scripts/, assets/) exist, so everything is inline in SKILL.md. Section headers are clear and navigation is easy, but the ~145-line entity list is exactly the kind of bulk reference material that should live in a separate file (or be cut), matching anchor 3 ("Some structure but could be better organized; ... content that should be separate is inline"). It is better than 2 (the body is not an unstructured wall — headers are logical) but not 4 since a large block of misplaced content dominates the file. | 3 / 5 |
Total | 12 / 20 Passed |