loop-coding

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.70). These links point to an external, third‑party skill marketplace (skills.sh) and its search/download endpoints (and specific skill pages) which the scripts in this repo download and execute content from; because skills can include arbitrary scripts/binaries from an untrusted source, these URLs are potentially risky as malware distribution vectors.

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

• Third-party content exposure detected (medium risk: 0.30). Phase 1 research runs scripts/github-search.sh (via gh search ...) with an outsider-controlled research question from scripts/phase-1-research.sh, and the resulting repository metadata/README snippet text is then merged into RESEARCH.md for LLM ingestion during later Audit/Plan/Review phases.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill fetches the external skills catalog at runtime (e.g., https://skills.sh/api/search and related skills.sh URLs) and then downloads/rents external skill packages which can include arbitrary code executed or used by the agent, so runtime fetching from https://skills.sh/ is a high-confidence runtime dependency.