A CLI personal-assistant agent, built stage by stage to learn tools and MCP
Two tools with minimal documentation. calculator has a basic description and schema, but ask_researcher lacks parameter descriptions and output schema documentation. Neither tool has error handling guidance or recovery hints. Naming is reasonable (verb-noun pattern), but descriptions are sparse (calculator: 95 chars, ask_researcher: 138 chars, both below the 194-char baseline). No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite both being read-only. Parameter descriptions are present but generic. No pagination, no output schema documentation, no error classification.
Delegate a question to a researcher sub-agent that can read files and search notes. Use this for questions that require digging through the project's files or the saved notes. Returns the researcher's findings.
Evaluate a basic arithmetic expression. Use this for any math instead of computing it yourself.
ask_researcher lacks output schema documentation. LLMs cannot plan downstream actions or extract structured data from researcher findings.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite both tools being read-only. Agents cannot determine retry safety or side-effect risk.
calculator description lacks context on when to use it vs manual computation. No guidance on supported operators, precision limits, or error cases (e.g., division by zero).
ask_researcher parameter 'question' has no format constraints, length limits, or examples of good vs bad questions. LLMs may pass vague or malformed queries.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 49 | 2026-07-28+ | v2 |
No error handling guidance. If calculator receives invalid syntax or ask_researcher fails to find information, LLMs have no recovery path.