glass_do

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

The description richly discloses behavioral traits beyond the annotations. It explains that `settle` waits for the screen to stop changing between steps, the optional `then` runs after all actions succeed (with order settle → diff → screenshot), and it fails fast: if an action errors it reports which index failed and how many ran. It also notes that partial sequences may already have landed, which is critical for the agent to understand side effects. It explicitly rejects `type`'s `return` observe inside actions, and clarifies that `then` is text-only unless screenshot/diff image. This is substantial behavioral context that the annotations (readOnlyHint=false, destructiveHint=false) do not provide, and it does not contradict them.

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 well-structured and appropriately sized for a complex tool. It front-loads the core purpose (collapsing round-trips) and then progressively provides details on actions, settle, then, and fail-fast behavior. Every sentence adds value: the example use cases (login, form-fill, menu→item) are concise and illustrative, and the warning about not batching when you need to see a result is crucial. There is no redundant content, and the structure (core purpose → details → usage guidelines) is logical and efficient.

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?

Given the tool's complexity (nested actions, optional `then` with settle/diff/screenshot), the description is remarkably complete. It covers the key behaviors an agent needs: the action types, settle behavior, `then` semantics, fail-fast error reporting, and the restriction on `type`'s `return`. The output schema is not present, but the description implicitly explains what the tool returns (e.g., which index failed and how many ran, text-only results unless screenshot/diff image). The annotations provide minimal safety info (readOnlyHint=false, destructiveHint=false), but the description compensates for the lack of output schema and gives enough context for correct invocation. No critical gaps are apparent.

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?

The schema has 100% description coverage for the two top-level parameters (`actions` and `then`), and the nested action types are thoroughly documented. The description adds value by summarizing the structure of `actions` (list of objects with action field) and the optional `then`'s fields, but the detailed parameter semantics (e.g., per-action fields, `then`'s order) are already in the schema. The description does not add additional meaning beyond what the schema provides, but it reinforces key points like the rejection of `type`'s `return` in actions. Since schema coverage is 100%, baseline is 3, but the description's clarification of the `then` order and the `type` restriction nudges it to a 4.

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 the tool's purpose: runs an ordered sequence of input actions in one call, collapsing round-trips, with optional observation. It specifies the verb (run), resource (input actions), and mode (ordered sequence). It also distinguishes from siblings by explicitly listing the action types (click, move, drag, scroll, type, key, settle) and the use cases (known sequences like login, form-fill). While it doesn't name a sibling directly, it contrasts with the alternative of calling individual tools, making the unique purpose clear.

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 provides explicit guidance on when to use this tool: for KNOWN sequences (login, form-fill, menu→item) and when not to use it (if you must see a result to choose the next action, don't batch that part). It also mentions that `type`'s `return` observe is rejected inside actions, guiding the agent to use `settle` or `then` instead. While it doesn't name specific alternative tools like `glass_click` or `glass_type`, it clearly contrasts with the per-action round-trip approach, which is sufficient. No explicit exclusion of other tools, but the conditional guidance is strong.

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