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

Annotations only provide the flag set (destructive, non-idempotent, non-open-world); the description adds substantial behavior they do not encode, notably the mandatory two-phase preview/confirm protocol, ICP merge/override semantics, defaults seeding when signals are omitted, and the LinkedIn-URL-to-slug conversion with a user-verification step. It does not explain why the tool is flagged destructive, but the additive behavioral disclosure is unusually rich.

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?

Front-loaded correctly: purpose and the two-phase rule come first, then ICP merge semantics, then signal semantics. Every sentence carries information, but it runs to roughly eight dense sentences in a single block with no visual grouping, making key rules (like the empty-array clearing convention) easy to skim past.

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 nine-parameter tool with nested signal objects and no output schema, the description covers the genuinely hard parts: the merge/inheritance model, the confirmation flow, and signal construction including the keyword_tracking example. It stops short of explaining what the phase-1 preview returns or how confirmation_token/max_signals are meant to be used in practice.

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 already 100%, so a 3 is the baseline; the description nonetheless adds real meaning, such as icp_filters overriding only supplied keys while keeping job_titles/mandatory_keywords/excluded_companies/company_sizes, excluded_companies accepting employer names or LinkedIn URLs and storing slug names, and the required field set for each signal. max_signals, type, app_name, icp_description and linkedin_account_ids get no description-level elaboration.

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?

States a specific verb, resource and scope: 'Create a new draft agent under the request's scoped app.' The 'draft' qualifier and the two-phase flow distinguish it from siblings like draft_icp_edit, enable_agent, and update_icp without the agent needing to open another schema.

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?

Explicit operational rules cover the whole decision space: call twice (preview then confirm), omit icp_filters on a vague request to inherit the saved ICP, pass only the keys the user asked to change, pass an empty array to clear a list, omit signals to seed from ICP, use [] for no signals. It also instructs the agent to ask for missing details rather than fabricate.

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