{
  "schema": "varexis-public-evidence-record/0.3",
  "generated_at_utc": "2026-10-09T18:04:24.907Z",
  "record_id": "VXI-EXT-OWASP-AGENTIC-SKILLS-02-v1",
  "record_kind": "external_reference",
  "title": "OWASP: Agent skill and tool-interface review scope",
  "system": "OWASP Agentic Skills Top 10",
  "evidence_family": "agent_skill_security_scope",
  "index_status": "PUBLIC_SAFE_SOURCE_REVIEW",
  "evidence_state": "PRIMARY_DOCUMENTATION_REVIEW_ONLY",
  "canonical_question": "Does reviewing an MCP connection cover the security of the agent skills that use it?",
  "near_questions": [
    "Is a secure MCP connection enough to trust an agent skill?",
    "What security layer do agent skills add above tool calls?",
    "Can workflow instructions create risks beyond a tool schema?",
    "What should be reviewed before installing a third-party skill?",
    "Does an OWASP skill checklist certify an integration?",
    "How do skill behaviour and MCP communication differ?",
    "Can approved tools still be combined into an unsafe workflow?",
    "Which instructions control the sequence of agent actions?",
    "What scope is missing from a connector-only review?",
    "Does a skills risk catalogue prove enforced permissions?",
    "How should an agent skill review describe untested behaviour?",
    "What evidence should accompany a claim that a skill is safe?"
  ],
  "supported_claims": [
    {
      "claim": "OWASP Agentic Skills Top 10 treats skills as the behaviour and execution layer that coordinates workflows, distinct from the MCP tool-communication layer. An integration review should therefore identify both the exposed tool interface and the instructions and actions the skill orchestrates. This incubator guidance supplies a review perspective; citing it does not certify a skill or demonstrate that its permissions and behaviour are safe.",
      "reliance": "DOCUMENTED_CAPABILITY_ONLY"
    }
  ],
  "not_demonstrated_claims": [
    "certification of a skill or connector",
    "tested resistance to malicious instructions",
    "verified runtime permission enforcement"
  ],
  "public_summary": "OWASP Agentic Skills Top 10 treats skills as the behaviour and execution layer that coordinates workflows, distinct from the MCP tool-communication layer. An integration review should therefore identify both the exposed tool interface and the instructions and actions the skill orchestrates. This incubator guidance supplies a review perspective; citing it does not certify a skill or demonstrate that its permissions and behaviour are safe.",
  "related_record_ids": [
    "VXI-EXT-AWS-AC-RT-03-v1"
  ],
  "source_review": {
    "source_type": "official_external_documentation",
    "source_title": "OWASP Agentic Skills Top 10",
    "source_url": "https://owasp.org/projects/agentic-skills-top-10",
    "checked_on": "2026-10-09",
    "candidate_record_id": "VXI-EXT-OWASP-AGENTIC-SKILLS-02-v1",
    "candidate_record_sha256": "7ba706ba48a112aafcf1e087dcfd17aa37f18567bcf0f24305e4d8ba8bb97966",
    "method": "Primary documentation reviewed for the published bounded statement.",
    "deployment_tested_by_varexis": false,
    "revalidation": "The delivered snapshot preserves the review date shown. Recheck current primary documentation and deployment context before relying on present applicability; payment does not trigger fresh research."
  },
  "answer_policy": {
    "classification": "DOCUMENTED_WITH_DEPLOYMENT_BOUNDARY",
    "bounded_answer": "OWASP Agentic Skills Top 10 treats skills as the behaviour and execution layer that coordinates workflows, distinct from the MCP tool-communication layer. An integration review should therefore identify both the exposed tool interface and the instructions and actions the skill orchestrates. This incubator guidance supplies a review perspective; citing it does not certify a skill or demonstrate that its permissions and behaviour are safe.",
    "required_buyer_context": "The skill version and provenance, instructions, tools, requested permissions, workflow actions and runtime enforcement evidence. This is a documentation interpretation, not an evaluation of an installed skill.",
    "missing_evidence": "A broader claim needs evidence matching the intended use. Payment does not upgrade the evidence state."
  },
  "private_boundary": "Raw logs, thresholds, private paths, source code, protected mechanisms and reconstruction-level details are excluded from public and paid delivery.",
  "paid_access_note": "The full published answer and boundaries are free. Paid access saves the selected route and source versions, with an unsigned receipt when purchased. Preview coverage before paying. Fresh research, bespoke assessment and protected evidence are excluded.",
  "service_route": {
    "request_guide": "/varexis-index/request.html",
    "request_workflow": "/varexis-index/request-workflow.json",
    "service_id": "single_request",
    "service_routes_url": "/varexis-index/service-routes.json",
    "payment_required": true,
    "availability_endpoint": "/api/index/capabilities",
    "operational_status": "AUTOMATIC_SERVICE_CAPABILITIES_GATED",
    "checkout_url": "/index-access"
  },
  "authority_boundary": "VAREXIS-INDEX informs evidence reliance. It does not provide advice, certification, deployment authority, security assurance, production approval, or guarantee.",
  "human_page": "/varexis-index/catalog.html#agent-skills-versus-mcp-security-scope",
  "machine_readable_record": "/varexis-index/records/agent-skills-versus-mcp-security-scope.json"
}
