{
  "slug": "skillsec-eval-for-agent-skill-security-713987",
  "title": "SkillSec-Eval for Agent Skill Security",
  "dek": "SkillSec-Eval evaluates reusable LLM-agent skills across repository admission, semantic retrieval, planner selection, execution, and evolution.",
  "summary": "SkillSec-Eval evaluates reusable LLM-agent skill security across five lifecycle stages using 327 real-world skills.",
  "tags": [
    "SkillSec-Eval",
    "llm-agents",
    "reusable-skills",
    "agent-security",
    "skill-lifecycle",
    "threat-taxonomy",
    "security-evaluation",
    "prompt-injection"
  ],
  "published_at": "2026-07-18T22:07:13.178+00:00",
  "grade": 8.5,
  "agent_utility": 8.2,
  "price_usdc": 0.226,
  "stats": {
    "claims": 10,
    "entities": 14,
    "faq": 7,
    "downloads": 0,
    "paid_downloads": 0
  },
  "preview": {
    "faq_questions": [
      "What is SkillSec-Eval?",
      "Which lifecycle stages does the paper cover?",
      "Is the paper only about prompt injection?",
      "How many skills were evaluated?",
      "What empirical finding is reported?",
      "Does the abstract describe operational controls?",
      "Can this abstract support a BibTeX citation?"
    ],
    "entity_names": [
      {
        "name": "SkillSec-Eval",
        "type": "framework"
      },
      {
        "name": "Reusable skills",
        "type": "concept"
      },
      {
        "name": "Large Language Model agents",
        "type": "system"
      },
      {
        "name": "Skill lifecycle",
        "type": "concept"
      },
      {
        "name": "Threat taxonomy",
        "type": "method"
      },
      {
        "name": "Repository admission",
        "type": "lifecycle stage"
      },
      {
        "name": "Semantic retrieval",
        "type": "lifecycle stage"
      },
      {
        "name": "Planner selection",
        "type": "lifecycle stage"
      },
      {
        "name": "Execution",
        "type": "lifecycle stage"
      },
      {
        "name": "Skill evolution",
        "type": "lifecycle stage"
      },
      {
        "name": "Repository of 327 real-world skills",
        "type": "dataset"
      },
      {
        "name": "Prompt injection",
        "type": "security topic"
      },
      {
        "name": "Runtime execution",
        "type": "security topic"
      },
      {
        "name": "Security risks throughout the broader skill lifecycle",
        "type": "problem"
      }
    ],
    "related_work_titles": [
      "Prompt injection security research",
      "Runtime execution security research",
      "Reusable-skill security benchmarking"
    ],
    "application_industries": [
      "AI platform security",
      "Enterprise automation",
      "Security benchmarking",
      "AI governance"
    ],
    "glossary_terms": [
      "Reusable skill",
      "LLM agent",
      "Skill lifecycle",
      "Threat taxonomy",
      "Repository admission",
      "Semantic retrieval",
      "Planner selection",
      "Execution",
      "Skill evolution",
      "Prompt injection",
      "Security evaluation"
    ]
  },
  "purchase": {
    "protocol": "x402",
    "version": 2,
    "network": "eip155:8453",
    "asset": "USDC",
    "price": 0.226,
    "full_markdown_url": "https://x402.disruptive-concepts.com/api/public/article/skillsec-eval-for-agent-skill-security-713987/full",
    "full_json_url": "https://x402.disruptive-concepts.com/api/public/article/skillsec-eval-for-agent-skill-security-713987/full?format=json"
  }
}