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

Annotations already declare readOnlyHint=true, so the safety profile is covered, but the description adds a genuinely valuable behavioral fact: 'Keep it for the person and send it back whole; GoldSeam keeps no copy,' telling the caller it is the sole custodian of the package. The clause 'Optionally record where to continue' is ambiguous — it could mean embedding the continuation point in the returned summary or persisting it, and the latter would sit awkwardly with readOnlyHint. Nothing is said about output size or cost of summarizing a large package.

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 instruction is front-loaded, which is good, but the body is a single ~120-word run-on sentence with deeply nested parentheticals ('each with its source, what prompted it, the statements supporting it with their affirmation...'), which is genuinely hard to parse. The information is substantive rather than filler, but the structure works against comprehension when it could have been a short bulleted list.

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?

An output schema exists, so return values need not be explained, and the description does cover the summary scope plus the important no-copy retention rule. Gaps remain relative to the tool's complexity: the 'note' parameter is unaddressed, the mutation-vs-read nature of 'record where to continue' is unclear, and there is no guidance on whether a full package is required in practice. Adequate but not complete for a nested-package tool.

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 67% and both the package and continue_with properties already carry descriptions, so the schema does most of the work. The description reinforces that the package must be sent back whole and that continue_with is optional, but adds no meaning for the undocumented 'note' field and no detail on continue_with's group/question_key structure beyond what the schema hints. Baseline 3 for a moderately-covered schema.

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 opening clause 'Summarize the interview so far' gives a specific verb and resource, and the long enumeration nails down exactly what the checkpoint contains (statements, interests, possibilities, awaiting answers, unread documents, continuation point). It also implicitly separates itself from show_me_my_selfie by noting that captured content 'is shown by show_me_my_selfie' rather than here. The density of the enumeration hurts readability but an agent can still tell what the tool produces.

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?

'Summarize the interview so far' and 'say what survives this chat ending' imply a checkpoint/end-of-session moment, which is usable context. However, it never states when to prefer this over siblings such as explore_from_my_selfie, respond_to_my_statements, or show_me_my_selfie, and gives no preconditions (e.g., whether a prior make_my_selfie answer is required). Usage is inferred rather than stated.

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