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

Annotations already declare destructiveHint=true and openWorldHint=true, which covers the safety profile. The description adds meaningful context beyond the annotations: the action is reversible ('Действие обратимо') but the time window with the wrong version on the address cannot be recovered. It also adds the important caveat that app platforms (ssr_next, streamlit, gradio, flask) may not support rollback at all — this is behavioral context not captured in annotations. Slight gap: doesn't detail auth permissions needed or rate limits, but the reversibility nuance and app-platform caveat add real value.

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?

Description is appropriately structured: a single clear purpose sentence, a bolded warning about side effects, reversibility caveat, and then a paragraph about app-platform limitations. Front-loaded with the most critical information (modifies what visitors see). The app-platform paragraph adds length but earns its place with critical behavioral nuance. Slightly long but well-organized with clear section breaks.

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?

This is a destructive, side-effect-heavy operation with no room for under-specification. The description covers: what it does, its reversibility, the irrecoverable consequence, when it fails (app platforms), the required user confirmation protocol, and fallback behavior (suggest rebuilding). There is an output schema present which handles return-value documentation. For a high-stakes mutation tool, this is thorough and complete.

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%, so both parameters are already well-documented in the schema: project is described as slug or id with guidance to pull slug from my_projects, and deploy is described as the build id with default behavior clarified. The description itself doesn't add parameter semantics beyond the schema, but the baseline of 3 is appropriate given complete schema coverage.

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?

Description clearly states the verb+resource: 'Вернуть на живой адрес предыдущую рабочую сборку' (return previous working build to live URL). The title 'Откатить сайт на прошлую сборку' and name 'rollback' are mutually reinforcing, and the description adds concrete behavioral detail about what gets changed (what visitors see). Clearly distinguishes from siblings like retry_deploy and cancel_deploy by specifying it switches to a previous working build on the live address.

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?

Excellent guidance. Explicitly says to ask the human before invoking unless rollback was explicitly requested ('спроси человека перед вызовом, если он не попросил откат прямо'). Clearly states it modifies what visitors see (a side effect requiring confirmation). Also gives explicit when-to-use-elsewhere guidance: for app platforms (ssr_next, streamlit, gradio, flask) rollback often doesn't work, and instructs the agent to explain failure to the user rather than retry. This is exemplary usage guidance.

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