skills/simota/agent-skills/sigil/Gen Agent Trust Hub

sigil

Pass

Audited by Gen Agent Trust Hub on Oct 22, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill inherently exposes an indirect prompt injection surface because its primary workflow involves scanning and processing external repository data (such as CLAUDE.md, package manifests, and source code) to inform the generation of new agent skills. If an attacker contributes malicious instructions to these project files, they could potentially influence the content of the generated skills, leading to unauthorized actions when those skills are executed by the user.
  • Ingestion points: Repository manifest files (package.json, pyproject.toml, etc.), configuration files, and rule files (CLAUDE.md, AGENTS.md) read during the SCAN and DISCOVER phases.
  • Boundary markers: While the skill emphasizes structure, it does not use unique delimiters or explicit 'ignore' instructions for the ingested repository content during the generation process.
  • Capability inventory: The skill possesses the ability to write and install executable skill files (.claude/skills and .agents/skills) and execute shell commands via project-local tools.
  • Sanitization: The skill documentation explicitly instructs the agent to escape XML-style angle brackets in frontmatter to prevent prompt injection and suggests using template patterns to separate static logic from dynamic user data.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 22, 2026, 01:03 PM