race-condition

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). The content is a high-risk, dual-use exploitation playbook: it gives step-by-step, tool-backed techniques (HTTP/1.1 last-byte gating, HTTP/2 single-packet bursts, Turbo Intruder scripts, h2spacex examples) explicitly aimed at abusing race conditions to commit fraud (double-claims, double-spend, inventory oversell, rate-limit bypass), although it does not contain backdoor code, covert exfiltration, credential-theft routines, obfuscated payloads, or supply-chain compromise artifacts.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.70). The skill explicitly directs sending and reading responses from arbitrary target web endpoints (e.g., "Capture the state-changing request in a proxy" and the parallel-request templates in §0, §2.1 and §5/Turbo Intruder), so the agent will fetch and interpret untrusted third‑party HTTP content as part of its workflow.