create-project-skills
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONPERSISTENCE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the repository being scanned, including manifest files, documentation, and source code. Maliciously crafted content within the repository could influence the pattern extraction process and the resulting instructions.\n
- Ingestion points: Project manifests (package.json, Cargo.toml, pyproject.toml, go.mod, etc.), instruction files (AGENTS.md, .cursorrules), and the entire source tree via rg and find utilities.\n
- Boundary markers: The skill does not employ boundary markers or specific instructions to ignore embedded prompt injection attempts in the source data.\n
- Capability inventory: The skill utilizes sub-agents via spawn_agent and leverages the $create-skill tool to persist findings.\n
- Sanitization: No sanitization or validation mechanisms are present to filter out potentially malicious instructions extracted from the codebase.\n- [PERSISTENCE]: The skill is designed to generate and save persistent instruction files (SKILL.md) in directories like .claude/skills/ or .agents/skills/. These files are automatically loaded in future agent sessions, enabling any potentially malicious or injected instructions to persist across the project's lifecycle.
Audit Metadata