Content
61%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 is a dense, well-structured operator's manual for the HubSpot toolkit: concrete tool slugs, parameter shapes, response quirks, and per-workflow sequences with role labels. Its main weaknesses are duplicated pitfall content across three sections, and the absence of post-batch validation or error-recovery feedback loops in batch-create workflows, which caps workflow clarity.
Suggestions
Add validation/feedback loops to the batch workflows: after HUBSPOT_CREATE_CONTACTS, verify the returned IDs (or re-search) to confirm record creation, and add 'if a batch fails or hits the 100-record limit, chunk and retry' guidance.
Deduplicate the pitfalls: keep one canonical 'Known Pitfalls' section (or per-workflow blocks) and drop the verbatim repeats of batch limits, internal property names, pagination, and duplicate-prevention.
Include one complete example payload for the most common call (e.g., a full filterGroups body for HUBSPOT_SEARCH_CONTACTS_BY_CRITERIA with valid operator values) to make the search guidance copy-paste ready.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly efficient — it assumes CRM/API knowledge and never explains what HubSpot or OAuth is — but the same pitfalls are repeated nearly verbatim across three places: per-workflow 'Pitfalls' blocks ('Max 100 per batch', 'Use paging.next.after'), the 'Known Pitfalls' section ('Batch limits: Max 100 records...', 'Cursor pagination: Use paging.next.after'), and again in 'Quick Reference' ('max 100'). Plus an ad-style footer ('*Powered by Composio*'). This matches 'mostly efficient but ... could be tightened' better than the level-4 'minor instances' anchor, though it is well above level 2 since there is no concept explanation or padding prose. | 3 / 5 |
Actionability | Concrete, executable guidance throughout: exact tool slugs (HUBSPOT_CREATE_CONTACTS), parameter shapes ('inputs array of {properties} objects, max 100 per batch', 'filterGroups array with {filters: [{propertyName, operator, value}]}'), response paths ('paging.next.after', 'response.data.results'), and a task-to-tool quick reference table. It stops short of the level-5 anchor because there are no complete example payloads (e.g., a full filterGroups body or a valid operator list), so a first call to HUBSPOT_SEARCH_CONTACTS_BY_CRITERIA still requires RUBE_SEARCH_TOOLS to fill in details. | 4 / 5 |
Workflow Clarity | Each workflow has a clear 'When to use' clause and a sequenced tool list with explicit roles (Prerequisite/Required/Optional/Alternative/Fallback), plus pre-flight checks ('Always search before creating to avoid duplicates', 'Run HUBSPOT_GET_ACCOUNT_INFO first'). However, the skill centers on batch operations (batch create contacts/companies, max 100) and includes no post-operation validation or error-recovery feedback loop — no verify-created-records step, no 'if the batch fails, chunk and retry' guidance. Per the rubric's cap, batch operations without validation/feedback loops cap workflow clarity at 3. | 3 / 5 |
Progressive Disclosure | The skill is a single self-contained file (no references/, scripts/, or assets/ exist, and the body has no broken file references) with well-organized sections: setup, five workflows, common patterns, known pitfalls, and a quick reference table. This matches 'Good structure; most content is appropriately placed; minor organization gaps' — the gap being that the per-workflow parameter/pitfall detail (~175 lines) could partly live in a reference file so SKILL.md stays an overview. It is not 5 because the content is not split across one-level-deep reference files, and not 3 because nothing is buried or poorly signaled. | 4 / 5 |
Total | 14 / 20 Passed |