ccf-research-engineer
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external, potentially untrusted codebases and project documentation, which is a standard surface for engineering-focused agents.
- Ingestion points: The skill reads and evaluates local repositories, project instructions, and documentation as defined in
references/codebase-and-implementation.md. - Boundary markers: The skill utilizes "family file contracts" and explicit resource limits defined in
SKILL.mdto scope its work, though it does not specify technical delimiters for untrusted data isolation. - Capability inventory: The skill possesses capabilities for shell command execution, file system modification, and system profiling using host-exposed tools.
- Sanitization: The skill relies on explicit instructions for the agent to manually inspect the codebase and environment before making changes, but no automated sanitization of external content is specified.
- [SAFE]: The skill includes explicit security and best-practice warnings, such as forbidding the exposure of credentials or private machine details in public artifacts and advising against broad upgrades or global package replacements.
Audit Metadata