ccf-idea-optimizer
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill's documentation in
references/source-notes.mdsuggests running a local maintenance script (python ..\ccf-common\scripts\check_sources.py) to verify the research source registry. This is a developer-facing maintenance command for the vendor's internal tooling. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted user-supplied research ideas and literature references, which presents an attack surface for indirect prompt injection.
- Ingestion points: User research seeds and literature papers are ingested through
references/idea-intake.mdandreferences/literature-grounded-evolution.md. - Boundary markers: The skill does not define specific delimiters or boundary markers when interpolating user-provided ideas into its development prompts.
- Capability inventory: The skill utilizes the
ccf-literature-searchertool to perform external network searches and retrieve literature. - Sanitization: While the instructions direct the agent to use
privacy-and-evidence.mdto ensure search terms are public-safe, there is no automated sanitization or filtering of the ingested user content.
Audit Metadata