Skip to main content
Glama

jev_verify

Read-only

Verify up to 100 claims against one evidence source, returning supported, contradicted or not_addressed with calibrated confidence before you assert them.

Instructions

Check up to 100 claims against one body of evidence: supported / contradicted / not_addressed per claim, with a calibrated confidence. Pass evidence_path for anything you have not already read — do NOT read a file in order to pass its text. The server reads it, chunks it if it is large, checks every claim against every chunk and returns line ranges. Use evidence only for text you already hold. Use it before you assert something to the user or write it into a file: hold your draft's claims to the source, or check a summary against the document it summarises. The rubric is strictly literal and closed-world: supported only if the evidence states or entails it. A claim that is true in the world but absent from the evidence comes back not_addressed — which is the answer you want when hunting unsupported assertions. conflicting means one part of the evidence firmly supports it and another firmly contradicts it. Claims should be single, self-contained statements — split compound sentences and resolve pronouns first. all_supported is true only when every claim is supported AND the model was confident about each; treat review/escalate gates as claims a human should look at.

Tool call template

Replace each placeholder with a value matching the input schema. Only required parameters are included; add optional parameters as needed.

{
  "id": 1,
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "arguments": {
      "claims": "<claims>"
    },
    "name": "jev_verify"
  }
}

Input Schema

NameRequiredDescriptionDefault
claimsYesSelf-contained statements to check, one per entry. Split compound claims.
end_lineNoWith `evidence_path`: last line to use, 1-based inclusive. Default end of file.
evidenceNoText you already hold. The only material the claims are judged against.
start_lineNoWith `evidence_path`: first line to use, 1-based inclusive. Default 1.
thresholdsNoOverride the server's configured gating thresholds for this call only.
evidence_pathNoA file for the server to read, relative to the project root. Preferred over pasting file text.

Output Schema

NameRequiredDescriptionDefault
modelYesThe versioned model id that actually answered.
usageYesToken usage. Jev charges for input tokens only.
claimsYesOne result per input claim, in input order.
summaryYesCounts across all claims.
latency_msYesWall-clock time for the underlying API call(s), including retries.
thresholdsYes
est_cost_usdNoPresent for file evidence: estimated input-token cost of this call.
all_supportedYesTrue only if every claim is `supported` and every gate is `auto`.
evidence_pathNoPresent when the evidence was read from a file; relative to the root.
evidence_chunksYesHow many requests the evidence was split into. 1 unless it was large.

Schema Changelog

Changes observed during successful MCP inspections.

  1. Changed13 schema fields changedv0.7.0
    • addedInput schema / properties / end_line
      Added value: +{
      +  "description": "With `evidence_path`: last line to use, 1-based inclusive. Default end of file.",
      +  "maximum": 9007199254740991,
      +  "minimum": 1,
      +  "type": "integer"
      +}
    • changedInput schema / properties / evidence / description
      Previous value: -"The only material the claims are judged against."New value: +"Text you already hold. The only material the claims are judged against."
    • addedInput schema / properties / evidence_path
      Added value: +{
      +  "description": "A file for the server to read, relative to the project root. Preferred over pasting file text.",
      +  "minLength": 1,
      +  "type": "string"
      +}
    • addedInput schema / properties / start_line
      Added value: +{
      +  "description": "With `evidence_path`: first line to use, 1-based inclusive. Default 1.",
      +  "maximum": 9007199254740991,
      +  "minimum": 1,
      +  "type": "integer"
      +}
    • changedInput schema / required
      Previous value: -[
      -  "claims",
      -  "evidence"
      -]New value: +[
      +  "claims"
      +]
    • changedOutput schema / properties / claims / items / properties / verdict / enum
      Previous value: -[
      -  "supported",
      -  "contradicted",
      -  "not_addressed"
      -]New value: +[
      +  "supported",
      +  "contradicted",
      +  "not_addressed",
      +  "conflicting"
      +]
    • addedOutput schema / properties / claims / items / properties / where
      Added value: +{
      +  "additionalProperties": false,
      +  "description": "Which lines of the evidence settled it. Present when the evidence came from a file or was chunked.",
      +  "properties": {
      +    "end_line": {
      +      "maximum": 9007199254740991,
      +      "minimum": -9007199254740991,
      +      "type": "integer"
      +    },
      +    "start_line": {
      +      "maximum": 9007199254740991,
      +      "minimum": -9007199254740991,
      +      "type": "integer"
      +    }
      +  },
      +  "required": [
      +    "start_line",
      +    "end_line"
      +  ],
      +  "type": "object"
      +}
    • addedOutput schema / properties / est_cost_usd
      Added value: +{
      +  "description": "Present for file evidence: estimated input-token cost of this call.",
      +  "type": "number"
      +}
    • addedOutput schema / properties / evidence_chunks
      Added value: +{
      +  "description": "How many requests the evidence was split into. 1 unless it was large.",
      +  "maximum": 9007199254740991,
      +  "minimum": -9007199254740991,
      +  "type": "integer"
      +}
    • addedOutput schema / properties / evidence_path
      Added value: +{
      +  "description": "Present when the evidence was read from a file; relative to the root.",
      +  "type": "string"
      +}
    • addedOutput schema / properties / summary / properties / conflicting
      Added value: +{
      +  "description": "Firmly supported by one part of the evidence and contradicted by another.",
      +  "maximum": 9007199254740991,
      +  "minimum": -9007199254740991,
      +  "type": "integer"
      +}
    • changedOutput schema / properties / summary / required
      Previous value: -[
      -  "supported",
      -  "contradicted",
      -  "not_addressed",
      -  "needs_review"
      -]New value: +[
      +  "supported",
      +  "contradicted",
      +  "not_addressed",
      +  "conflicting",
      +  "needs_review"
      +]
    • changedOutput schema / required
      Previous value: -[
      -  "claims",
      -  "summary",
      -  "all_supported",
      -  "thresholds",
      -  "model",
      -  "usage",
      -  "latency_ms"
      -]New value: +[
      +  "claims",
      +  "summary",
      +  "all_supported",
      +  "thresholds",
      +  "evidence_chunks",
      +  "model",
      +  "usage",
      +  "latency_ms"
      +]
  2. First observedv0.1.4

TDQS

A4.3/5.0
Behavior5/5

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

Goes well beyond readOnlyHint/openWorldHint: it discloses the closed-world literal rubric, that the server reads and chunks large files and returns line ranges, that a true-but-absent claim returns not_addressed, what 'conflicting' means, and the exact semantics of all_supported including review/escalate gating. These are non-obvious behaviors an agent could not infer from annotations or schema.

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 the core action and return shape, then use-cases, then rubric semantics, then gating. Slightly long at six paragraph blocks and the rubric paragraph restates the literal/closed-world idea twice, but nearly every sentence carries operational information.

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?

Despite six parameters including a nested thresholds object and file-path handling, the description covers input choice, evaluation semantics, output verdict vocabulary, and success-gate meaning. An output schema exists, yet the extra explanation of verdicts and all_supported is legitimately clarifying rather than redundant.

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

Parameters4/5

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

Schema coverage is 100%, so baseline is 3, but the description adds real decision value: evidence_path is preferred for anything unread and evidence is only for text already held, the 100-claim cap, and the requirement that claims be single self-contained statements with pronouns resolved. start_line/end_line and thresholds are left entirely to the schema.

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

Purpose4/5

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

"Check up to 100 claims against one body of evidence" is a precise verb+resource+scale statement, and the three verdicts plus 'calibrated confidence' tell the agent what comes back. It never names or contrasts the siblings (jev_evaluate, jev_rank, jev_pick, jev_gate_action), so the agent cannot rule them out from this text alone.

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

Usage Guidelines4/5

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

Gives concrete trigger conditions: "Use it before you assert something to the user or write it into a file: hold your draft's claims to the source, or check a summary against the document it summarises." It also gives a clear when-not-use-path rule (do not read a file yourself to pass its text). No sibling alternatives are offered, so it stops short of full routing guidance.

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