direct_method

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

Annotations already indicate readOnlyHint=true, so the description does not need to repeat safety. It adds valuable behavioral context: the estimator's dependence on correct Q-model specification, the extrapolation risk outside logged action support, and the recommended alternative under misspecification. It does not contradict the annotations, and the added failure-mode detail goes beyond what structured fields 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 description is well-structured with clear labels (Assumptions, Pre-conditions, Failure modes, Alternatives, Typical minimum N) and is front-loaded with the core purpose. Each section adds useful information, though the density is high and the 'Typical minimum N' could be seen as non-essential. Overall, it is appropriately sized for the tool's complexity.

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 12-parameter OPE tool, the description covers the method, assumptions, preconditions, failure modes, alternatives, and sample size guidance. However, it does not explain key parameters like pi_target (the target policy probabilities), n_actions, or how data is mapped to X/A/R, which an agent would need for correct invocation. The output schema exists but is not sufficient to fill these gaps.

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 100%, but most parameter descriptions are tautological ('A parameter (np.ndarray).', 'pi_target parameter.'). The tool description compensates by clarifying that X is context, A is logged action, and R is reward, which is meaningful semantic enrichment. However, pi_target and n_actions remain unexplained, so the description does not fully compensate for the weak schema descriptions.

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 clearly identifies the tool as 'Direct outcome regression (plug-in Q-model) OPE', which is specific enough for an expert to know it performs off-policy evaluation via a plug-in outcome model. It also names alternatives, distinguishing it from doubly robust, IPS, and SNIPS estimators. However, there is no explicit verb like 'estimate' or 'evaluate', relying on the OPE acronym to convey the action.

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 explicitly states assumptions (correct Q-model specification, no unmeasured confounding), pre-conditions (X, A, R available), and failure modes (model misspecification bias). It directly instructs the agent to prefer the doubly-robust estimator in the failure case and lists alternatives (sp.doubly_robust, sp.ips, sp.snips), giving clear when-to-use and when-not-to-use guidance.

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