working-with-skills
Pass
Audited by Gen Agent Trust Hub on Oct 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill functions exclusively as an instructional resource (a manual) for agents interacting with the PostHog Skill MCP tools. It does not contain any malicious code, hidden instructions, or attempts to bypass safety filters.
- [SAFE]: The guide advocates for robust security and data integrity practices, such as optimistic concurrency control (using base_version) to prevent data loss and progressive disclosure to minimize context pollution and potential injection surfaces.
- [INDIRECT_PROMPT_INJECTION]: The skill describes a system for managing external skill content, which inherently involves an indirect prompt injection attack surface. Ingestion points: The agent is instructed to fetch external skill data using skill-get and skill-file-get as documented in SKILL.md. Boundary markers: The skill emphasizes efficient fetching but does not specify delimiters or security wrappers for the retrieved content. Capability inventory: The agent possesses capabilities to create and modify files (skill-create, skill-update) and potentially execute shell commands (Bash) as noted in the documentation. Sanitization: The skill does not provide specific instructions for sanitizing or escaping the content of retrieved skills before the agent processes them.
Audit Metadata