Skip to main content
Glama
Sign In
caveat-lang

caveat-lang

by WSattazahn
OverviewSchemaSourceRelated ServersDiscussionsAdmin
Rust
Local

caveat_dependents

Read-onlyIdempotent

Find what rests on a named evidence item, reading stream, or caveat by replaying events, revealing impact before changing or withdrawing that input.

Instructions

Replay optional events and find what rests on one named evidence item, reading stream or caveat. Use for impact inspection; caveat_explain instead explains all decisions and their reasons. Subjects are causes only: evidence, a reading stream or a caveat. Claims and decisions are never subjects, because a claim causes nothing and the grounds of a decision are evidence, not claims; both appear in the report as what the subject reaches. To trace from a claim, query the evidence that supports or opposes it (report.claims lists those relations). report (caveat-dependents/0.1) includes subject, kind, sequence, event outcomes, claims, decisions, changes, values, displayed values and withdrawal records; dependency entries distinguish grounds from possible influence through lineage. exitCode: 0 replay completed (may include rejected events); 1 fatal event with last good state; 2 load failure or unknown subject, with report.error. Returns the same JSON in structuredContent and text: {schema:"caveat-authoring/0.1",operation,sourceSha256,runtime,exitCode,report}. Read exitCode even when isError is false. Bridge failures instead set isError:true with {schema:"caveat-authoring-error/0.1",error:{kind,message}}.

Input Schema

NameRequiredDescriptionDefault
eventsNoEvents replayed in array order from a fresh session. Omitted or [] inspects the initial state; previous calls supply no history. Rejections are recorded and replay continues; a fatal event stops replay at the last good snapshot. At most 1,000 entries and 1 MiB of serialized UTF-8 JSON.
sourceYesComplete inline Caveat program text, available as the virtual file inline.cav. No file path or URL lookup. At most 1 MiB of UTF-8 source; each call uses a fresh runtime.
subjectYesExact case-sensitive evidence name, reading stream name, caveat name, or evidence occurrence ID after replay. Evidence records supplied observations; a reading stream is a named history of samples; a caveat is a qualification on evidence. For example, probe selects its evidence/readings, values selects a stream, and values@1 selects one recorded reading. Use a name from source or caveat_explain; not a file path. A claim or decision name is refused like an unknown name, with exitCode 2.

Schema Changelog

Changes observed during successful MCP inspections.

  1. Changed1 schema field changedv0.1.2
    • changedInput schema / properties / subject / description
      Previous value: -"Exact case-sensitive evidence name, reading stream name, caveat name, or evidence occurrence ID after replay. Evidence records supplied observations; a reading stream is a named history of samples; a caveat is a qualification on evidence. For example, probe selects its evidence/readings, values selects a stream, and values@1 selects one recorded reading. Use a name from source or caveat_explain; not a file path, claim or decision name. Unknown names return exitCode 2."New value: +"Exact case-sensitive evidence name, reading stream name, caveat name, or evidence occurrence ID after replay. Evidence records supplied observations; a reading stream is a named history of samples; a caveat is a qualification on evidence. For example, probe selects its evidence/readings, values selects a stream, and values@1 selects one recorded reading. Use a name from source or caveat_explain; not a file path. A claim or decision name is refused like an unknown name, with exitCode 2."
  2. Changed6 schema fields changedv0.1.1
    • addedInput schema / properties / events / description
      Added value: +"Events replayed in array order from a fresh session. Omitted or [] inspects the initial state; previous calls supply no history. Rejections are recorded and replay continues; a fatal event stops replay at the last good snapshot. At most 1,000 entries and 1 MiB of serialized UTF-8 JSON."
    • addedInput schema / properties / events / items / description
      Added value: +"One event to dispatch to the supplied program, for example {\"event\":\"read\",\"payload\":{\"value\":85}} when source declares read with a value parameter."
    • addedInput schema / properties / events / items / properties / event / description
      Added value: +"Event name declared in source, such as read or assess. Supply the name only, without Caveat syntax."
    • addedInput schema / properties / events / items / properties / payload / description
      Added value: +"Values keyed by the event parameter names declared in source, for example {\"value\":85}. Omit for an empty payload ({}). The runtime checks parameter names, types and declared bounds."
    • changedInput schema / properties / source / description
      Previous value: -"Inline Caveat source; no file or URL lookup."New value: +"Complete inline Caveat program text, available as the virtual file inline.cav. No file path or URL lookup. At most 1 MiB of UTF-8 source; each call uses a fresh runtime."
    • addedInput schema / properties / subject / description
      Added value: +"Exact case-sensitive evidence name, reading stream name, caveat name, or evidence occurrence ID after replay. Evidence records supplied observations; a reading stream is a named history of samples; a caveat is a qualification on evidence. For example, probe selects its evidence/readings, values selects a stream, and values@1 selects one recorded reading. Use a name from source or caveat_explain; not a file path, claim or decision name. Unknown names return exitCode 2."
  3. First observedv0.1.0

TDQS

A4.6/5.0
Behavior5/5

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

Annotations already declare readOnly/idempotent/non-destructive, but the description adds non-obvious behavior: exitCode 0/1/2 meanings, rejected events recorded vs fatal events stopping at last good snapshot, 'Read exitCode even when isError is false,' and a separate error envelope for bridge failures. It also notes each call uses a fresh runtime with no history carried over.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness4/5

Is the description appropriately sized, front-loaded, and free of redundancy?

Front-loaded with purpose and alternative, and almost every sentence carries substantive routing or error-handling information. It is dense and long, with several nested clauses (subject rules, report contents, exit codes) that could be tightened, but little is pure filler.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

With no output schema, the description shoulders the burden and does so: it enumerates report contents (subject, kind, sequence, outcomes, claims, decisions, changes, values, withdrawals), the response envelope, and the exitCode/isError interplay. Nothing an agent needs to call or interpret it is missing.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters3/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema description coverage is 100% and the description's subject/events guidance (case-sensitive names, examples like values@1, refusal of claim/decision names) largely mirrors the schema. It adds the rationale that claims cause nothing, which is mild added value, so baseline 3 is the right level.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

Opens with a specific verb+resource+scope: 'Replay optional events and find what rests on one named evidence item, reading stream or caveat.' It immediately names the sibling it is not (caveat_explain) and states the input type, so an agent can distinguish it from the other caveat_* tools without opening a schema.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines5/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

Explicit routing: 'Use for impact inspection; caveat_explain instead explains all decisions and their reasons.' It also states when-not at the subject level ('Claims and decisions are never subjects') and gives a workaround ('To trace from a claim, query the evidence that supports or opposes it'). Clear when/when-not/alternative.

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

Deploy Server

Other Tools