Dispatch a Harness run

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

Annotations already declare destructiveHint=true, openWorldHint=true and readOnlyHint=false, but the description adds real behavioral context the annotations cannot: the scope fields are 'task instructions, not enforced per-path filesystem isolation,' and read content may be sent to the configured model provider even though Relay uses a loopback address. It also notes the returned session link is stable and that a terminal status (not webUrl) marks completion. It stops short of describing dispatch timing/blocking behavior or reuse semantics for an existing sessionId.

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 core action is front-loaded in the first sentence and the cautionary sentences are dense with operative constraints rather than filler. It loses a point because roughly half the description is a verbatim duplicate of the 'task' parameter's schema description, so an agent reading both is told the same thing twice.

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?

With 19 parameters, no output schema, and destructive/open-world annotations, the description covers what an agent most needs: what it returns (stable session link), how completion is judged (wait_run to terminal, then assistantText), the non-enforcement of scoping, and the data-egress caveat. It leaves gaps around session reuse behavior and the interaction between externalPaths, permissionPreset, and operator policy.

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 63%, and the description meaningfully disambiguates the scope parameters beyond their schema text: reviewTargets 'identify what to assess; they are not a read whitelist' while contextReadScope 'declares where Harness may search and read,' and writeScope must be empty for read-only runs. It adds no semantics for model, provider, reasoningEffort, agentPreset, sessionMode, idempotencyKey, or content, though several of those carry their own in-schema descriptions.

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 first sentence states a specific verb and resource chain — create/reuse a Harness session, select provider/model/reasoning/presets, submit the first task, return a session link — which is far more specific than the terse title 'Dispatch a Harness run'. It does not explicitly distinguish itself from siblings like queue_run, start_review, or steer_run, so an agent must infer the boundary from the queue/review naming alone.

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?

There is one genuinely useful routing instruction: 'Sharing webUrl is not completion: keep wait_run until a terminal status, then consume assistantText,' which names a sibling and the follow-up condition. However, it never says when to choose start_run over queue_run or steer_run, nor when sessionMode/sessionId reuse is appropriate, leaving the primary selection decision implied.

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