mail_send_reconcile

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

With zero annotations, the description must carry the full behavioral load, and it makes a strong start: a persistent global writer fence on begin, a refusal path for missing authority, and an explicit negative list of side effects (no send, no retry, no Mail wake, no page-ledger change, no retry permission). It still omits idempotency, what happens on a partially completed begin, and how long a fence persists.

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?

Five tight sentences, front-loaded with the core no-send guarantee and the two-action structure, with zero filler. The negative list at the end is dense but each clause adds distinct information. Slight density of undefined jargon is the only cost.

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 15-parameter, 8-required, two-phase protocol tool with no annotations and no output schema, the description leaves too much open: no explanation of the parameters, no statement of what begin or commit return, and no lifecycle semantics. It successfully conveys the safety envelope but not enough to invoke the tool correctly.

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 description coverage is only 13% across 15 parameters, so the description should compensate but names no parameter explicitly. Words like "evidence," "approval," and "fence" loosely gesture at evidence_ref, approval_ref, and fence_id, but the roles of claim_id, attempt, revision, fingerprint, fingerprint_algorithm, deployment_ref, outcome, and reason_code remain wholly unexplained.

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 specific verb phrase ("No-send claim adjudication") and then spells out the two action modes: begin erects a persistent global writer fence, commit verifies completion conditions. It also carves out a clear boundary against send-like siblings ("Neither action sends, retries, wakes Mail..."), so an agent can distinguish it from mail_send. The jargon ("claim adjudication," "page ledger") keeps it from 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?

Usage is implied through preconditions rather than stated: legacy claims are ineligible, commit needs administrator approval and complete Sent AND Outbox evidence, and missing authority refuses. An agent can infer eligibility, but there is no explicit "use this instead of mail_send when..." routing nor mention of what to do after a refusal. No sibling alternatives are named.

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