Act on the screen

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

Annotations only declare readOnly=false, destructive=false, idempotent=false, openWorld=false; the description adds real behavior beyond that — every call performs one step, then blocks until the screen settles and returns verdicts (changed/settled), snapshot quality, health, and new elements. The implicit retry hazard of a non-idempotent mutation is not spelled out, and failure/error semantics are only hinted at via 'health'.

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 with the action set and the post-action contract, then targets, then the batching alternative, then a doc pointer — a sensible priority order with no filler. The semicolon-dense second sentence packs action-target pairing tightly enough that it takes a re-read, costing a point.

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 16-parameter tool with nested target/for/untilVisible objects, the description covers the essentials: what a call does, where targets come from, the settle-and-report contract, and the batching alternative. An output schema exists so return details need not be repeated, and the doc anchor covers the rest; only the session lifecycle (sessionId origin) is left implicit.

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 88% schema description coverage the schema already carries most parameter meaning, but the description adds the target-resolution model ('@eN ref from qa_snapshot, text, id, a native selector (WDA iOS), or x/y'), which the schema only enumerates field-by-field. It also flags the platform caveat that native selectors are WDA/iOS, information the schema does not convey.

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?

Opens with a concrete verb list (tap, type, clear, swipe, scroll, press, open_url, wait) applied to a single UI step, then states what happens after (waits for settle, reports changed/settled, snapshot quality, health, new elements). This is far more specific than the title 'Act on the screen' and lets an agent distinguish it from qa_snapshot (observation) and qa_flow_run (multi-step batching).

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?

Explicitly routes multi-step work away: 'For 3+ known steps, batch them in one qa_flow_run {flowYaml}' — a named alternative plus the condition that selects it. It also names qa_snapshot as the source of @eN refs, establishing the natural preceding tool. It stops short of an explicit 'when not to use' (e.g. for reading state without acting, use qa_snapshot), so not a full 5.

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