{
  "schema_version": 1,
  "builder_plan": {
    "scenario": "An agent reads external documents or tool responses that may contain malicious instructions.",
    "change": "Replay attacks and legitimate lookalikes in an isolated copy of your agent workflow.",
    "baseline": "Your existing checks with the same tasks, permissions, and tool responses.",
    "steps": [
      "Locate the paper's implementation and check its license, model support, data needs, and hardware requirements. Stop here if you cannot obtain the required artifacts.",
      "Save a small set of representative product tasks and expected outcomes. Keep evaluation tasks out of any training or tuning, and choose quality and cost limits before running.",
      "Run the baseline and the proposed change on identical inputs. Keep other settings fixed and repeat paired runs to expose variation.",
      "Save per-task outputs, failures, timings, and all usage costs. Review regressions before expanding the test."
    ],
    "measure": "Successful attacks missed, legitimate actions blocked, detection delay, and review cost.",
    "decision": "Keep the change only if it meets your preselected quality and latency limits and improves the metric you care about after setup and operating costs. Otherwise retain the baseline. This is an exploration plan, not a validated recommendation."
  },
  "builder_review": null,
  "arxiv_id": "2609.31039",
  "title": "MetaPermit: Scalable and Auditable Access Control for AI Agents via LLM-Inferred Meta-Attributes",
  "summary": "MetaPermit replaces ad-hoc combinations of static permission rules and direct LLM judgments on each proposed tool call with a decoupled design: an LLM infers a compact set of task-independent meta-attributes per call, and a fixed deterministic policy authorizes or denies, yielding auditable decisions. It costs an additional LLM inference pass per tool call (latency and API spend), plus the engineering effort of defining and maintaining the meta-attribute schema and policy rules. It can fail if the inference model is misled by indirect prompt injection embedded in the analyzed context, if meta-attribute extraction is noisy, or if the fixed policy is incomplete for edge cases the schema does not capture.",
  "rationale": "The method runs on open-weight models and small components compatible with a single-GPU or API-only setup, and it directly addresses a production-relevant risk for tool-using agents, but its claimed robustness gains must be validated against the reader's specific tools, threats, and latency budget before adoption.",
  "technique": "Policy-based agent tool authorization via LLM-inferred meta-attributes",
  "family": "seguranca_e_guardrails",
  "recommendation": "testar",
  "published": "2026-09-25",
  "as_of": "2026-10-18",
  "editorial_status": "indexed",
  "source_url": "https://arxiv.org/abs/2609.31039",
  "independent_implementations": 0,
  "total_implementations": 1,
  "stars": 1,
  "citations": null,
  "claims": [
    {
      "claim_id": "claim-01",
      "statement": "MetaPermit replaces ad-hoc combinations of static permission rules and direct LLM judgments on each proposed tool call with a decoupled design: an LLM infers a compact set of task-independent meta-attributes per call, and a fixed deterministic policy authorizes or denies, yielding auditable decisions. It costs an additional LLM inference pass per tool call (latency and API spend), plus the engineering effort of defining and maintaining the meta-attribute schema and policy rules. It can fail if the inference model is misled by indirect prompt injection embedded in the analyzed context, if meta-attribute extraction is noisy, or if the fixed policy is incomplete for edge cases the schema does not capture.",
      "basis": "inferred",
      "result": "",
      "baseline": "",
      "conditions": "",
      "source_url": "https://arxiv.org/abs/2609.31039",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "claim_id": "claim-02",
      "statement": "MetaPermit produces 31% more consistent authorization decisions than LLM-driven authorization, improves task completion over CaMeL and IPIGuard by up to 109%, and reports no malicious tool calls executed on AgentDojo and AgentDyn.",
      "basis": "inferred",
      "result": "",
      "baseline": "",
      "conditions": "",
      "source_url": "https://arxiv.org/abs/2609.31039",
      "source_page": null,
      "source_excerpt": ""
    }
  ],
  "exposure_map": [
    {
      "dimension": "quality",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "dimension": "compute",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "dimension": "latency",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "dimension": "operations",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "dimension": "compatibility",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "dimension": "security",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "dimension": "data_and_training",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "dimension": "reproducibility",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    }
  ],
  "risks": [],
  "minimum_test": [],
  "open_questions": [],
  "independent_tests": [],
  "lab_reproductions": [],
  "report_available": false,
  "validation_tier": "unknown",
  "evidence_tier": "unknown",
  "infrastructure_basis": "unknown",
  "equations": [],
  "equations_status": "not_fetched",
  "core_kind": "",
  "equations_fetched_at": "",
  "equations_source_url": "",
  "equations_source_sha256": "",
  "exposures_reviewed_at": "",
  "exposures_source_sha256": ""
}