Skip to main content
Glama

veritas_adversary_gate

Stress-tests claims against adversarial attack transforms like bound inflation, evidence removal, and perturbation. Returns PASS or MODEL_BOUND verdict with fragility score for final robustness check.

Instructions

Gate 9/10: Stress-tests the claim against attack transforms (bound inflation, evidence removal, parameter/evidence perturbation). Use this as the final robustness check; fragility > 25% triggers MODEL_BOUND (ADVERSARY_FRAGILE). Returns JSON with verdict (PASS | MODEL_BOUND), fragility (float), attacks_tested (int), attacks_degraded (int).

Input Schema

TableJSON Schema
NameRequiredDescriptionDefault
claimYesA VERITAS BuildClaim object for deterministic gate evaluation. All fields are optional for partial evaluation — only fields relevant to the invoked gate are required.

Schema Changelog

Changes observed during successful MCP inspections.

  1. Added
  2. Removedv2.1.1
  3. Changed16 schema fields changedv1.0.2
    • changedInput schema / properties / claim / description
      Previous value: -"A VERITAS BuildClaim object containing all declared primitives, operators, regimes, boundaries, loss models, evidence items, cost vectors, and policy configuration needed for deterministic gate evaluation. All fields are optional for partial evaluation — only the fields relevant to the gate being invoked are required."New value: +"A VERITAS BuildClaim object for deterministic gate evaluation. All fields are optional for partial evaluation — only fields relevant to the invoked gate are required."
    • changedInput schema / properties / claim / properties / attack_suite / description
      Previous value: -"AttackSuite with suite_id and list of Attack transforms (InflateBound, RemoveEvidence, PerturbParam, PerturbEvidence) for adversary gate."New value: +"Attack transforms for adversary testing: InflateBound, RemoveEvidence, PerturbParam, PerturbEvidence."
    • changedInput schema / properties / claim / properties / boundaries / description
      Previous value: -"Named boundary constraints (e.g. 'CAPEX_USD <= 100000') that the claim must satisfy."New value: +"Boundary constraints, e.g. {name: 'cap', constraint: 'CAPEX_USD <= 100000'}."
    • changedInput schema / properties / claim / properties / commit / description
      Previous value: -"Git commit SHA for reproducibility and audit trail."New value: +"Git commit SHA for reproducibility."
    • changedInput schema / properties / claim / properties / cost / description
      Previous value: -"CostVector with optional fields: compute_flops, memory_bytes, wall_clock_s, capital_usd, coordination_agents."New value: +"CostVector: compute_flops, memory_bytes, wall_clock_s, capital_usd, coordination_agents."
    • changedInput schema / properties / claim / properties / cost_bounds / description
      Previous value: -"Upper bounds for each cost component. Utilization = max(cost_i / bound_i). Values must be > 0."New value: +"Upper bounds for each cost component. All values must be > 0."
    • changedInput schema / properties / claim / properties / dependencies / description
      Previous value: -"SBOM-style dependency manifest for supply-chain analysis: package names, versions, registries, and integrity hashes."New value: +"SBOM-style dependency manifest: package names, versions, registries, hashes."
    • changedInput schema / properties / claim / properties / evidence / description
      Previous value: -"Evidence items, each with id, variable, value (Numeric or Categorical), timestamp, method, provenance, and optional dependencies."New value: +"Evidence items with id, variable, value, timestamp, method, provenance."
    • changedInput schema / properties / claim / properties / loss_models / description
      Previous value: -"Named loss functions as ArithmeticExpr over primitives, with optional upper bounds."New value: +"Loss functions as ArithmeticExpr with optional upper bounds."
    • changedInput schema / properties / claim / properties / operators / description
      Previous value: -"Declared operators with name, arity, input primitive names, output primitive name, and totality flag."New value: +"Operators with name, arity, input/output primitive names, totality flag."
    • changedInput schema / properties / claim / properties / policy / description
      Previous value: -"PolicyConfig overrides: hash_alg, solver_backend, timeouts, thresholds. Defaults to VERITAS Omega v1.3.1 canonical values if omitted."New value: +"PolicyConfig overrides for hash_alg, solver_backend, timeouts, thresholds."
    • changedInput schema / properties / claim / properties / primitives / description
      Previous value: -"Declared typed variables with name, domain (Interval/EnumSet/FiniteSet), optional units, and description."New value: +"Typed variables with name, domain (Interval/EnumSet/FiniteSet), optional units."
    • changedInput schema / properties / claim / properties / project / description
      Previous value: -"Unique project identifier, e.g. 'omega-brain-mcp'."New value: +"Project identifier, e.g. 'omega-brain-mcp'."
    • changedInput schema / properties / claim / properties / regimes / description
      Previous value: -"Named operating regimes, each with a predicate ConstraintExpr over declared primitives."New value: +"Operating regimes with name and predicate ConstraintExpr."
    • changedInput schema / properties / claim / properties / security / description
      Previous value: -"Security posture declaration: SAST results, secret scan findings, injection surfaces, auth boundaries, and TLS/crypto configuration."New value: +"Security posture: SAST results, secret scan, injection surfaces, auth, TLS config."
    • changedInput schema / properties / claim / properties / version / description
      Previous value: -"Semantic version string of the build being evaluated, e.g. '2.1.0'."New value: +"Semantic version, e.g. '2.1.0'."
  4. Changed16 schema fields changedv1.0.1
    • changedInput schema / properties / claim / description
      Previous value: -"Full or partial VERITAS BuildClaim object"New value: +"A VERITAS BuildClaim object containing all declared primitives, operators, regimes, boundaries, loss models, evidence items, cost vectors, and policy configuration needed for deterministic gate evaluation. All fields are optional for partial evaluation — only the fields relevant to the gate being invoked are required."
    • addedInput schema / properties / claim / properties / attack_suite / description
      Added value: +"AttackSuite with suite_id and list of Attack transforms (InflateBound, RemoveEvidence, PerturbParam, PerturbEvidence) for adversary gate."
    • addedInput schema / properties / claim / properties / boundaries / description
      Added value: +"Named boundary constraints (e.g. 'CAPEX_USD <= 100000') that the claim must satisfy."
    • addedInput schema / properties / claim / properties / commit / description
      Added value: +"Git commit SHA for reproducibility and audit trail."
    • addedInput schema / properties / claim / properties / cost / description
      Added value: +"CostVector with optional fields: compute_flops, memory_bytes, wall_clock_s, capital_usd, coordination_agents."
    • addedInput schema / properties / claim / properties / cost_bounds / description
      Added value: +"Upper bounds for each cost component. Utilization = max(cost_i / bound_i). Values must be > 0."
    • addedInput schema / properties / claim / properties / dependencies / description
      Added value: +"SBOM-style dependency manifest for supply-chain analysis: package names, versions, registries, and integrity hashes."
    • addedInput schema / properties / claim / properties / evidence / description
      Added value: +"Evidence items, each with id, variable, value (Numeric or Categorical), timestamp, method, provenance, and optional dependencies."
    • addedInput schema / properties / claim / properties / loss_models / description
      Added value: +"Named loss functions as ArithmeticExpr over primitives, with optional upper bounds."
    • addedInput schema / properties / claim / properties / operators / description
      Added value: +"Declared operators with name, arity, input primitive names, output primitive name, and totality flag."
    • addedInput schema / properties / claim / properties / policy / description
      Added value: +"PolicyConfig overrides: hash_alg, solver_backend, timeouts, thresholds. Defaults to VERITAS Omega v1.3.1 canonical values if omitted."
    • addedInput schema / properties / claim / properties / primitives / description
      Added value: +"Declared typed variables with name, domain (Interval/EnumSet/FiniteSet), optional units, and description."
    • addedInput schema / properties / claim / properties / project / description
      Added value: +"Unique project identifier, e.g. 'omega-brain-mcp'."
    • addedInput schema / properties / claim / properties / regimes / description
      Added value: +"Named operating regimes, each with a predicate ConstraintExpr over declared primitives."
    • addedInput schema / properties / claim / properties / security / description
      Added value: +"Security posture declaration: SAST results, secret scan findings, injection surfaces, auth boundaries, and TLS/crypto configuration."
    • addedInput schema / properties / claim / properties / version / description
      Added value: +"Semantic version string of the build being evaluated, e.g. '2.1.0'."
  5. First observedv1.0.0

TDQS

A4.2/5.0
Behavior4/5

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

Returns JSON with verdict, fragility, attacks_tested, attacks_degraded. Discloses threshold trigger. No annotations provided, so description adequately covers behavioral aspects.

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

Conciseness5/5

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

Two sentences: first defines purpose, second specifies usage and output. No filler, front-loaded with key information.

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

Completeness4/5

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

Covers purpose, usage, output format. Lacks detailed explanation of attack transforms but overall complete for a gate tool given input schema richness.

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%; the description does not add extra parameter-specific meaning beyond what schema already provides. Baseline 3 applies.

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?

Description clearly states the tool 'stress-tests the claim against attack transforms' and labels it as the 'final robustness check', distinguishing it from sibling gate tools. The verb+resource combo is specific.

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?

Explicitly says 'Use this as the final robustness check' and specifies the fragility threshold. No explicit exclusions or alternatives named, but the context implies it's the last step.

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