supervisor-api
Replace the client-side agent loop with Databricks Supervisor API (hosted tools). Use when: (1) User asks about Supervisor API, (2) User wants Databricks to run the agent loop server-side, (3) Connecting Genie spaces, UC functions, agent endpoints, or MCP servers as hosted tools.
90
88%
Does it follow best practices?
Impact
Pending
No eval scenarios have been run
Advisory
Suggest reviewing before use
Quality
Discovery
100%Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.
This is a strong skill description that clearly defines a specific capability (replacing client-side agent loops with Databricks Supervisor API), includes an explicit 'Use when' clause with three well-defined trigger scenarios, and uses domain-specific terminology that makes it highly distinguishable from other skills. The description is concise yet comprehensive, covering both the what and when effectively.
| Dimension | Reasoning | Score |
|---|---|---|
Specificity | Lists specific concrete actions: replacing client-side agent loop with Supervisor API, running agent loop server-side, and connecting specific tool types (Genie spaces, UC functions, agent endpoints, MCP servers) as hosted tools. | 3 / 3 |
Completeness | Clearly answers both 'what' (replace client-side agent loop with Databricks Supervisor API) and 'when' (explicit 'Use when:' clause with three numbered trigger scenarios). | 3 / 3 |
Trigger Term Quality | Includes strong natural trigger terms users would say: 'Supervisor API', 'Databricks', 'agent loop', 'server-side', 'hosted tools', 'Genie spaces', 'UC functions', 'agent endpoints', 'MCP servers'. These cover the domain-specific vocabulary well. | 3 / 3 |
Distinctiveness Conflict Risk | Highly distinctive with a clear niche around Databricks Supervisor API and hosted tools. The specific terminology (Genie spaces, UC functions, Supervisor API) makes it very unlikely to conflict with other skills. | 3 / 3 |
Total | 12 / 12 Passed |
Implementation
77%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This is a strong, highly actionable skill with clear step-by-step workflow and excellent concrete examples covering all tool types, permissions, tracing, and the MCP approval flow. Its main weakness is redundancy — the tracing setup code and explanation appear twice, inflating the document significantly. With deduplication and better progressive disclosure (moving tracing details to a separate file), this would be excellent.
Suggestions
Remove the duplicated tracing code — the full `_get_trace_destination()` function and tracing explanation in the 'Enabling Tracing' section repeats what's already in Step 3. Keep it in Step 3 and make 'Enabling Tracing' a brief reference or move detailed tracing docs to a separate file.
Consider extracting the 'MCP Server Tools: Multi-Turn Approval Flow' and 'Enabling Tracing' sections into separate reference files to reduce the main skill's length and improve progressive disclosure.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The skill is mostly efficient but has significant redundancy — the tracing code and `_get_trace_destination()` function appear in full twice (once in Step 3 and again in the 'Enabling Tracing' section). The 'How It Works' explanation and some contextual notes could be tightened. The supported models table and limitations list are useful but the overall document could be ~30% shorter. | 2 / 3 |
Actionability | Excellent actionability — provides fully executable Python code for the agent server, concrete tool declaration examples for all five tool types, specific shell commands for testing/deploying, a permissions table with exact permission names, and a concrete example of the MCP approval flow input structure. Everything is copy-paste ready with clear placeholders. | 3 / 3 |
Workflow Clarity | The 5-step workflow is clearly sequenced (install → declare tools → update agent → grant permissions → test/deploy) with each step building on the previous. The MCP multi-turn approval flow is explicitly documented with numbered steps. Troubleshooting section provides error-recovery guidance. The tracing section includes a validation check (verify MLFLOW_EXPERIMENT_ID) with user-facing prompts for missing values. | 3 / 3 |
Progressive Disclosure | The skill references other skills ('add-tools' skill, 'supervisor-api-background-mode' skill) and external docs appropriately, but the document itself is quite long (~250 lines) with the tracing content duplicated inline rather than split to a reference file. The 'Enabling Tracing' section repeats code already shown in Step 3 and could be a separate reference. No bundle files are provided to offload detail. | 2 / 3 |
Total | 10 / 12 Passed |
Validation
100%Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.
Validation — 11 / 11 Passed
Validation for skill structure
No warnings or errors.
- Repository
- databricks/app-templates
- Commit
dfeb4ac
- Reviewed
Table of Contents
Is this your skill?
If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.