Forget (state transition, never silent deletion)

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

With no annotations provided, the description carries the full burden of behavioral disclosure. It clearly states that archive is recoverable, deprecate preserves the memory for opt-in reads, and purge is the only irreversible deletion (subject to git history as a backstop). It also emphasizes 'never silent deletion,' setting expectations that all state changes are explicit. It does not cover error handling or idempotency, but the key behaviors are disclosed well.

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 description is a single, dense sentence organized by mode with colons and semicolons. Each mode's effect and prerequisites are front-loaded, with no filler. The nuclear phrase 'never silent deletion' in the title signals non-obviousness, and every clause earns its place. It is concise yet fully informative, ideal for quick agent parsing.

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 with 5 parameters, an enum, and no output schema, the description covers the mode semantics, prerequisites, and recovery guarantees. It does not detail the return format (irrelevant without an output schema) or mention error conditions, but it fully equips an agent to invoke the tool correctly per the described behavior. Minor gaps like id format or store path patterns are not critical to judging correctness of the call.

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 60% (store, reason, and confirm have descriptions, id and mode do not descriptively). The description adds substantial meaning: it explains what each mode means and the conditional requirements (reason required for deprecate, confirm must be true for purge). This goes beyond the schema, which merely provides enum values and defaults. The id parameter's meaning is left implicit, but since it's the only required field and the tool's purpose centers on existing memories, the omission is minor.

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 title 'Forget (state transition, never silent deletion)' prefaces the tool as a state transition rather than a deletion, and the description enumerates three distinct modes (archive, deprecate, purge) with clear effects. This precisely distinguishes it from siblings like memory_update (which would modify content) and memory_recall (which reads), so an agent can immediately understand what this tool does and how it differs from other memory operations.

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?

The description provides explicit usage scenarios for each mode: archive for recoverable removal from default recall, deprecate for judged-wrong memories with a required reason, and purge as the only true deletion, restricted to archived/deprecated states with confirm=true. It does not explicitly name sibling tools as alternatives, but the mode-specific guidance makes the when-to-use decision clear. Missing an explicit 'do not use for simple edits' pointer, but the context is implicit.

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