explain_recipe_delta

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

Annotations already declare readOnlyHint=true and the description reinforces it ('without modifying either file'), then adds material context the annotations can't carry: hard limits (50,000 cells, 64 columns, 10,000 result cells per source), exact-decimal numeric evidence, the distinction between numeric_adjustment and business explanation, and the clarification that explanation_token is an integrity fingerprint, not authorization or a commit token.

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 core action, then organized into prerequisites, outputs, optional-parameter behavior, pinning, and limits. It is dense for a 10-parameter tool and several sentences pack two or three clauses with semicolons, but almost nothing is filler.

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 complex tool with no output schema it still describes the return shape (added/removed/updated contributions, group changes, bounded source-row examples, omitted contributions reported) and enumerates operational limits. The remaining gap is that the required `recipe` object — a free-form nested structure — has no field-level guidance anywhere, which is a real risk given 20% schema coverage.

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 only 20%, so the description must compensate, and it does for nearly every parameter: key_fields must be exact headers present and unique on every row, metric selects one exact output alias, group must supply all group_by fields with exact literal values, sample_limit is bounded 1-10 and only affects examples, and the expected_*_sha256 parameters are described via the pin-from-previous-response guidance. Only the shape of the nested recipe object remains unspecified.

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 plus scope: 'Explain sum/count changes by source identity, without modifying either file' and 'Recompute one shared Recipe v1 on two distinct .xlsx source versions.' It implicitly separates itself from diff_workbooks and run_workbook_recipe by requiring two source versions plus a shared recipe, but it never names an alternative explicitly.

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?

Gives clear preconditions (nonblank unique key_fields on every row, at most 8 sum/count metrics, sha256 pinning from a previous response for repeatable drilldown) and an explicit exclusion list: does not validate saved report cells, infer historical recipes, evaluate formulas, or establish business causes. It stops short of naming a sibling to use instead when those exclusions apply.

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