Inspect Stack Frame

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

Beyond the readOnlyHint/openWorldHint annotations, the description discloses substantial behavior: the response is trimmed by default and full: true lifts the trim; lazy values may need expansion; excludeStepValues resolves lazy values before comparison so neutral lazy factors are dropped; withHighlights returns an A1-keyed overlay plus raw grid; and a terminal session answers 409. There is no contradiction with the annotations — 'read' aligns with readOnlyHint=true, and openWorldHint aligns with the caveat that the world may have moved on.

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 single dense paragraph is information-rich with no filler — every clause covers a distinct behavior (frame contents, laziness, trimming, filters, highlights, 409 semantics). The core purpose is front-loaded and the terminal-session caveat is placed at the end. At roughly 180 words for a tool with 7 params and decision-table complexity it is appropriately sized, though the wall-of-text structure with nested parentheticals could be more scannable.

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?

With no output schema, the description carries the burden of explaining return values — and it does, enumerating input parameters, runtime context, result, sub-steps, decision, ruleNames, and the optional highlight overlay. It covers preconditions (suspended only), error semantics (409), parameter behaviors, and the openl_get_trace_value handoff for lazy values. Nothing an agent needs to call this tool correctly is 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?

Schema coverage is 100%, so the baseline is 3, and the description genuinely adds meaning beyond the schema: it explains the 'why' of the filters (surfacing an anomaly), the interplay between onlyExecutedSteps and excludeStepValues, and the lazy-value resolution behavior before neutral comparison. It also frames the decision/ruleNames output as the 'killer feature' for per-rule breakpoints, giving purpose beyond the schema's bare property definitions.

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 names a specific verb and resource — 'Freeze and read the full state of one suspended stack frame' — and enumerates exactly what that state contains (input parameters, runtime context, result, sub-steps with computed values, decision/ruleNames). It clearly distinguishes itself from siblings like openl_get_trace_value (lazy expansion) and other trace tools. An agent can tell exactly what this tool returns without opening the 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 an explicit precondition and exclusion: 'Valid only while suspended (a terminal session answers 409 — read its final state from the last returned stack)' tells the agent when not to call and what to do instead. It also routes lazy-value expansion to openl_get_trace_value and explains when the step filters are warranted (surfacing an anomaly among neutral factors). It stops short of contrasting with every related trace sibling (e.g., openl_expand_trace_tree, openl_watch_trace_cells), so the guidance is strong but not exhaustive.

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