Single tool 'start_job' has a clear action verb and reasonable description (95 chars), but lacks critical details. Input schema is present with two parameters (label, seconds), both typed and described. However, the description references '_meta' and 'tasks/get' without explaining what those are or how the agent should use them. No output schema is documented. The tool modifies state (starts a job) but the description does not explicitly state this is a write operation or explain idempotency/retry semantics. Parameter 'seconds' lacks min/max bounds despite the description mentioning '1 - 120', this constraint should be in the schema, not just prose. Overall: functional but incomplete for production LLM use.

• Add explicit output schema documenting the response structure: e.g. { task_id: string, status: string, created_at: ISO8601, ... }. Include the _meta field structure.
• Move 'seconds' constraint into JSON Schema: add 'minimum': 1, 'maximum': 120 to the parameter definition.
• Expand description to state: 'Starts a background job that runs for the specified duration. This is a write operation; results are idempotent if called with the same label. Poll the returned task_id using tasks/get to check completion.'
• Clarify the 'label' parameter: add description like 'A human-readable name for this job (max 100 chars). Used for identification in logs and UI.'
• Document error cases: what happens if seconds is out of range? If label is empty? If the job board is full? Provide actionable recovery guidance.

Inferred effective spec: 2026-07-28+.

• Current-spec patterns present: stateless request handling, Streamable HTTP