subagent_start

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

With no annotations, the description carries the full burden and does substantial work: it discloses the background/async nature, the immediate job_id return (or inline result under wait_seconds), inherited account/plugins/MCP context, and that the host enforces account limits with a suggested action on refusal. It leaves gaps on timeout expiry behavior and resource consumption, but the return contract and account-limiting semantics are genuinely useful additions.

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?

Four dense sentences, front-loaded with what the tool does and its return contract, then the task-authoring guidance. Every sentence carries information, though the enumeration of shared resources ('MCP servers, plugins, skills and settings') is slightly padded.

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?

For a 12-parameter launch tool with no annotations and no output schema, the description covers the essentials: it explains the return value (job_id or inline result), the background execution model, and the failure path. It is missing only ancillary behaviors like what happens at timeout_seconds or how the delegated job is monitored, which the sibling tools imply.

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 coverage is 67% and the schema itself already documents the meaningful parameters (access trade-offs, resume_job, wait_seconds, cwd, add_dirs). The description only indirectly touches account ('under one of your accounts') and adds no syntax or semantics for model, effort, runtime, max_turns or timeout_seconds, so baseline 3 is appropriate.

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 gives a specific verb and resource ('Start a subagent') and immediately clarifies what that means operationally: another installed agent CLI runs one task in the background with shared MCP servers, plugins and settings. It doesn't explicitly name sibling tools to contrast with (subagent_wait, subagent_result), but no sibling could be mistaken for this launcher.

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?

It gives real task-authoring guidance ('complete, self-contained task: goal, scope, files, and how to check the result') and hints at the wait_seconds trade-off, plus what to do if a start is refused. However it never says when to use this versus kicking off and polling via the sibling tools, or when resume_job is the right entry point instead.

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