The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill evaluates research ideas and manuscripts provided by the user or retrieved from external sources, which creates a surface for indirect prompt injection where malicious instructions could be embedded in the content being reviewed.
Ingestion points: Ingests user-supplied research ideas and external academic literature fetched via the literature searcher tool.
Boundary markers: The instructions mandate a "concept-only review" boundary and refer to report templates to structure output, but do not employ strict delimiters for the raw input content.
Capability inventory: The skill is authorized to write to the project-relative ccfa-workfiles/ directory and can transition to other specialist skills like ccf-idea-optimizer and ccf-paper-reviewer.
Sanitization: Instructions specify following privacy-and-evidence.md to prevent leaking private idea details during grounding searches.
[COMMAND_EXECUTION]: The documentation includes instructions for validating academic sources by running a local Python script.
Evidence: python ..\ccf-common\scripts\check_sources.py in references/source-notes.md.
Context: The command targets a script within the sibling project directory ccf-common, representing a local developmental tool execution.

ccf-idea-reviewer