kai-social
Pass
Audited by Gen Agent Trust Hub on Aug 22, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill performs automated exploration of project files including README.md, package.json, and generic project files to extract marketing context without user intervention. This ingestion of untrusted data lacks boundary markers or sanitization, creating a surface where malicious instructions embedded in project files could influence the agent's behavior.
- Ingestion points: MARKETING.md, CLAUDE.md, README.md, package.json, landing pages, and any discovered project files.
- Boundary markers: Absent; the skill does not utilize specific delimiters to isolate external project data from instructions.
- Capability inventory: File writing to the workspace/social/ directory and the project root (MARKETING.md).
- Sanitization: Absent; no mention of filtering or validating content extracted from the codebase.
- [DATA_EXPOSURE_AND_EXFILTRATION]: The automated discovery process explicitly targets configuration files related to email, ads, and analytics to extract monetization and strategy details. While this information is intended for the local MARKETING.md file, the broad scan increases the risk of extracting and potentially exposing sensitive configuration metadata.
Audit Metadata