kb-organize
Pass
Audited by Gen Agent Trust Hub on Apr 25, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by extracting metadata from markdown files and interpolating it into the project's CLAUDE.md configuration file.
- Ingestion points: Processes all
.mdfiles in thedocs/kb/directory recursively to read frontmatter (SKILL.md, Steps 2 and 3). - Boundary markers: Lacks explicit delimiters or instructions to the agent that the 'tags' and 'scope' content being written to
CLAUDE.mdis from an untrusted source. - Capability inventory: Possesses file write capabilities for
CLAUDE.md,_index.md, and_log.md, as well as directory creation and file movement (SKILL.md, Steps 5 and 6). - Sanitization: No validation or sanitization is performed on metadata strings before they are appended to the project's central configuration documentation.
Audit Metadata