start_ralph

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

The annotations include readOnlyHint: true and idempotentHint: true, which together suggest the tool is safe and non-destructive. However, the description describes initiating a multi-round cycle that likely makes changes (e.g., '每轮真实验证' implies code modifications and test runs). This contradicts the readOnlyHint, which should indicate a read-only operation. The description does not disclose behavioral traits beyond what annotations provide; it adds that the tool does not create background processes and that safe stop does not equal success, but these are minor. The contradiction between description and annotation is significant, warranting a score of 2.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Is the description appropriately sized, front-loaded, and free of redundancy?

The description is concise at roughly two sentences. It front-loads the purpose and key behavioral traits. However, the structure could be improved by separating the 'returns' section from the 'not' statements for clarity. There is no redundancy, and each clause serves a purpose, earning a 4.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

The tool has 14 parameters with 100% schema coverage, no output schema, and no nested objects. The description provides a high-level overview of what the tool does and its return values (Delegated Plan, Heartbeat evidence, optional front-end scripts). However, given the complexity of the tool (a multi-round development orchestrator), more context about the return format, how to use the evidence, and how failures are reported would be valuable. The presence of annotation contradictions further reduces completeness, making a 3 appropriate.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema description coverage is 100%, meaning every parameter already has a description in the schema. The tool description does not add any parameter-specific semantics beyond what the schema provides. However, there are 14 parameters, many with default behaviors tied to mode, and the description does not summarize key relationships (e.g., how mode affects max_rounds, confirm_every, etc.). The baseline for 100% coverage is 3, but given the complexity and the fact that the description could have provided a high-level overview of parameter interactions, a 4 is given because the schema descriptions are detailed and the tool description's lack of additional parameter guidance is not severely detrimental.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Does the description clearly state what the tool does and how it differs from similar tools?

The description clearly states that the tool is for long tasks requiring multi-turn small-step implementation with real verification and formal convergence. It uses specific verbs like 'start_ralph' implying initiation of a process, and contrasts with siblings by mentioning '多轮小步实现、每轮真实验证和正式收敛' which distinguishes it from simpler tools like start_feature or start_bugfix. However, the description is in Chinese, which might reduce clarity for non-Chinese speakers, and the title 'Ralph 循环开发编排' provides additional context.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Does the description explain when to use this tool, when not to, or what alternatives exist?

The description explicitly states the tool is for '长任务' (long tasks) and mentions what it does not do: '不自动运行循环、不创建后台进程' and '安全停止不等于成功'. This gives implicit guidance on when to use it (complex, iterative tasks) and when not (if you need automatic loops or background processes). However, it does not explicitly name alternatives or provide exclusion criteria relative to the listed sibling tools like 'plan_heartbeat' or 'resume_plan', so a 4 is appropriate.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.