Two tools with complete input schemas and descriptions, but both suffer from weak naming (job_output, job_kill lack action clarity), minimal parameter descriptions, and no documented output schemas. Descriptions are adequate (60-80 chars) but lack context on when to use each tool or what data structures are returned. No error handling guidance. Parameters are typed but lack constraints (e.g., max_output minimum is 256 but no maximum). Tool composition is reasonable (separate concerns) but naming does not follow verb_noun convention clearly.

• Rename 'job_output' to 'get_job_output' and 'job_kill' to 'cancel_job' to follow verb_noun convention and improve LLM intent inference.
• Document the output schema explicitly: job_output returns {status: string, output: string | null, file_path: string | null, elapsed_seconds: number}. Clarify when output is inlined vs written to file.
• Add descriptions to all parameters: 'reason' in job_kill should explain what this string is used for (logging, user notification, etc.). 'max_output' should document the file-writing side effect and when to use it.
• Add error handling guidance: 'If job_id not found, returns error code JOB_NOT_FOUND. If wait=true times out, returns TIMEOUT_EXCEEDED, retry with wait=false to poll manually. If job crashed, status field is FAILED with error details in output.'
• Expand tool descriptions to 100-150 chars: explain the polling vs blocking trade-off, mention the 300s timeout, and clarify that job_kill is idempotent (safe to retry).