act

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

Beyond the annotations (destructive/idempotent/openWorld), the description discloses that execution halts on first error, that the response carries a freshly read screen plus per-step verdicts (✓ verified, ~ sent but unverified, ✗ failure/refusal), that non-ASCII text is pasted via clipboard and echoed back verbatim to the user, and the wait timeout bounds (5s default, 30s max). These are exactly the side-effect and verification details an agent needs before committing a destructive action.

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 purpose and stopping rule are front-loaded, and nearly every clause carries unique information; the step list and the reporting legend both earn their space. It is, however, one dense run-on with stacked parentheticals that is harder to parse than it needs to be.

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 10-step, destructive, open-world tool with no output schema, the definition covers ordering, per-step field semantics, confirmation policy, and the shape of the return (verdicts + refreshed screen). It is only incomplete on the `then` output selector and never states the prerequisite of reading the screen with `look` before choosing targets.

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?

With reported schema coverage at 0%, the description does the heavy lifting by mapping each `do` value to its fields: `type` (target = field to touch first), `key` (key list plus the confirm rule for enter/KEYCODE_*), `scroll` (direction = reading direction, down to continue), `swipe` (target → to), `wait` (text appears, or gone=true), `open_url` (text = address). The notable gap is the top-level `then` parameter (both/image/elements/none), which is never explained and directly governs what the agent gets back.

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 opening sentence gives a specific verb+resource (gestures on the phone, executed in order, then the refreshed screen) and the enumeration of ten step types makes the scope concrete. It never explicitly contrasts itself with siblings such as `launch` or `settings` for the cases where those would be the right pick, so it falls just short of a 5.

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 operational rules: steps run in order and execution stops at the first error, `confirm=true` is required for enter and other KEYCODE_* and only on explicit user instruction or after a "yes", and targets are resolved against the screen read just before the gesture. What is missing is a stated alternative or when-not-to-use (e.g. "use look first", "use launch to start an app").

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