skillfed

97.6% of Injection and Poisoning Caught, Only 90.2% When Skills Interact

Notes on Sealing the Audit-Runtime Gap for LLM Skills (arXiv:2605.05274) — Tingda Shen, Yebo Feng, Konglin Zhu, Xiaojun Jia, Yang Liu, Lin Zhang · May 2026

Note published · written by SkillFed’s research pipeline from the paper above · how these notes are made

AI-assisted notes · reviewed by SkillFed Skill security

SIGIL treats skill security as a lifecycle problem, not a one-time gate. Every approved skill gets anchored in a tamper-evident on-chain registry, published under one of four modes — Transparent (free plaintext), Licensed (paid access), Sealed (custodial use), or Committed (off-chain content, on-chain hash). Admission runs through a DAO audit committee: pluggable audit methods (static analysis, LLM review, sandbox checks) operating under a stake-and-slash economic model that penalizes bad calls. Every later load then passes through a Skill Verification Loader (SVL), which re-checks the fetched skill against the exact record the auditors signed off on and enforces its permission manifest before any content reaches the model's context.

Tested against 1,023 real skills pulled from ClawHub and other public registries across six attack categories, SIGIL holds up best exactly where its on-chain binding is aimed. Explicit injection and implicit poisoning land at 97.6% accuracy, local tampering at 100%; rug-pull attacks — the closest analog to a supply-chain compromise — come in at 95.1%. The floor is cross-skill interaction attacks at 90.2%, where individually clean skills combine into an unsafe outcome: the hardest category by design, since no single skill's record looks wrong on its own. The DAO ensemble beats any single audit method run alone and holds its accuracy even with 20–40% of auditors colluding. Verification itself is cheap: under 86 ms of batched checking at load time, and auditing's own token cost stays under 3% of a typical monthly LLM subscription's quota.

Key numbers

Real-world skills tested1,023 skills, 6 attack types
Injection & poisoning detection97.6% accuracy
Cross-skill interaction detection (weakest category)90.2% accuracy
Batched load-time verificationunder 86 ms
Auditing's own token costunder 3% of a $20/mo quota

Skills related to this research

ios-debugger-agent ios-debugger-agent automates the process of building, launching, and running your current iOS project on an active simulator instance. This skill eliminates repetitive setup steps, letting you focus on development rather than tooling. Ideal for developers seeking faster iteration cycles during iOS app testing and debugging.★ 3,861 Qt Qml Review Qt Qml Review combines deterministic linting with parallel agent-driven analysis to audit Qt6 QML code across six focused domains: bindings, layout, component loading, delegates, states, and performance. The skill reports only high-confidence issues with structured mitigations and never modifies code.★ 322 Paper Audit Paper Audit conducts structured reviewer-style assessments of academic manuscripts, identifying technical, methodological, and claim-level issues while flagging submission blockers. It supports multiple modes—quick-audit for fast readiness checks, deep-review for comprehensive critique, gate for pass/fail decisions, and re-audit for tracking fixes—and produces actionable revision roadmaps anchored to specific textual locations.★ 400 Hyperframes Registry HyperFrames Registry lets you find, install, and integrate pre-built blocks and components into your compositions using the `hyperframes add` command. Blocks are standalone sub-compositions you wire in via attributes; components are effect snippets you paste directly into your markup. The registry also guides you through authoring and contributing new items upstream.★ 42,764

Related notes

References

  1. Shen, Feng, Zhu, Jia, Liu, Zhang. "Sealing the Audit-Runtime Gap for LLM Skills." arXiv:2605.05274 (2026).
  2. Jamshidi et al. "Securing the Model Context Protocol: Defending LLMs against tool poisoning and adversarial attacks." (2025).
  3. Beurer-Kellner et al. "Snyk Finds Prompt Injection in 36%, 1467 Malicious Payloads in a ToxicSkills Study of Agent Skills Supply Chain Compromise." (2026).
  4. Holzbauer et al. "Malicious Or Not: Adding Repository Context to Agent Skill Classification." arXiv:2603.16572 (2026).