sensemakr

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

The annotations only provide readOnlyHint=true, which is consistent with a sensitivity analysis tool. The description adds significant behavioral context: it discloses the modeling assumption (OLS of y on treat plus controls), defines the Robustness Value and its variants (rv_q, rv_qa), explains benchmarking, and details failure modes. This is valuable beyond the annotation. However, it does not mention whether it writes to disk or performs other side effects, but since readOnlyHint is set, the agent can safely assume it is a read operation. The description exceeds the baseline by providing framework details but could still mention whether it caches results (though as_handle parameter suggests it may), so a perfect 5 is not given.

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 structured with clear labels (Validation, Assumptions, Pre-conditions, Failure modes, Alternatives) and is front-loaded with the core purpose. It is long but every sentence adds information; no fluff. It could be slightly more concise, but the structure aids readability and the density of content justifies its length. A 4 reflects that it is well-organized but not extremely short.

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?

The tool is complex: it has 11 parameters, no output schema (has output schema true in signals, but not provided), and many sibling tools that perform similar analyses. The description covers the method's assumptions, key concepts, pre-conditions, failure modes, and alternatives. It even mentions suggested minimum N (50). Given the complexity and the lack of an output schema shown, the description is nearly exhaustive for an agent to know when and how to call it correctly. The detail parameter is self-explanatory with examples. No critical gap is apparent.

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 100% and each parameter has a description, so the baseline is 3. The tool description adds extra semantic context: it explains what 'benchmark' means (covariates to benchmark confounding strength against, and clarifies that benchmark names must be a subset of controls), and it also mentions rv_q and rv_qa outputs that relate to the alpha parameter. It also explains the 'detail' parameter's purpose (e.g., 'agent' for planning). This goes beyond the schema descriptions, so a 4 is appropriate.

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 begins with a specific verb phrase 'Sensitivity analysis for omitted variable bias', naming the resource (the sensitivity analysis) and the method (Cinelli & Hazlett 2020). It clearly distinguishes itself from siblings like oster_bounds and evalue that also do sensitivity analysis, by spelling out the partial-R-squared framework, Robustness Value, and benchmarking. The purpose is unambiguous and not a tautology.

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 names alternatives (sp.oster_bounds, sp.evalue, sp.unified_sensitivity) and states assumptions and pre-conditions (need treatment, outcome, controls). It also lists failure modes with corrective actions (e.g., benchmark names must be subset of controls, collinear controls cause instability and should be dropped). This tells the agent exactly when to use this tool over siblings and how to avoid common pitfalls.

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