replica_restart

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

The description discloses behavior the annotations cannot express: the tool is a deliberate no-op that 'refuses with a teaching error,' and dry_run previews that refusal honestly. There is mild tension with destructiveHint=true / idempotentHint=false, which reflect the nominal write intent rather than actual effects, but the description resolves that ambiguity rather than creating it, so this is added value, not a contradiction.

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 [WRITE][risk=high] marker and the single most important fact, then rationale, alternatives, and args. Slight redundancy in stating REST unavailability twice ('NOT available over Ray's REST API' and 'exposes no per-replica restart endpoint over REST'), but the repetition is arguably deliberate emphasis for a refusal tool.

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 tool that can never succeed, the description covers everything an agent needs: why it fails, what the error is, that dry_run mirrors the failure, and what to do instead. With no output schema and only nominal annotations, nothing material is left unexplained.

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 0% (bare titles only), so the description carries the burden and largely does: it defines application, deployment, replica_id with its provenance ('from replica_list'), dry_run's effect, and target's default-omission semantics. It does not restate types, but the meaning of each parameter is clear.

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?

States a specific verb and resource (restart one Serve replica) and immediately negates it with the crucial scope fact: it is NOT available over Ray's REST API. This lets an agent distinguish it from siblings like drain_replica, scale_replicas_up/down, and deployment_redeploy without opening any schema.

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 says when not to use it (no per-replica REST endpoint exists; the controller respawns unhealthy replicas itself) and names concrete alternatives ('scale the deployment down then up, or redeploy'). It even explains that dry_run returns the same unavailability rather than a false positive, closing the obvious loophole an agent would try.

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