horowitz_manski

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

Annotations only declare readOnlyHint and openWorldHint, so the description carries the behavioral burden, and it pays it well. It discloses that the output is an interval rather than a point, that bounds can fail by being too wide, that a credible identifying restriction or binary selection/attrition indicator is required, and it even notes typical minimum N and a validation tier—all beyond what annotations provide.

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 text is dense but well organized with labeled sections: Validation, Assumptions, Pre-conditions, Failure modes, Alternatives, and Typical minimum N. The method sentence is front-loaded, though small ambiguities like 'validated evidence tier' and the 'sp.' prefix on alternatives keep it from being maximally crisp.

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 14-parameter estimator with an output schema, the description supplies the non-obvious invocation context: identifying assumptions, interval nature, preconditions, failure mode, tightening strategy, alternatives, and sample-size caution. Nothing essential for correct selection and invocation appears to be missing.

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?

The schema already describes all 14 parameters, so the baseline is 3. The description adds only general context about covariates and a selection/attrition indicator for the Lee-bounds alternative; it does not add per-parameter semantics for alpha, n_boot, y_lower/y_upper, as_handle, or result_id beyond what the schema already states.

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 description opens with 'Horowitz-Manski (2000) bounds conditioning on covariates,' which identifies the method and its conditioning role, and later clarifies that the result is an interval under weak assumptions. However, it never explicitly names the estimand being bounded (e.g., ATE/ATT), so the agent must infer the precise object from the method name and parameter 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?

The description gives selection-relevant contrast: weak set-identifying assumptions versus Lee bounds' monotonicity and Oster's proportional-selection assumption, plus a named Alternatives list with sp.oster_delta, sp.lee_bounds, and sp.manski_bounds. It stops short of an explicit if-then routing rule like 'use this when only weak assumptions are credible,' but the assumption comparisons give a clear context for choosing among siblings.

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