kai-funnel-audit
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [COMMAND_EXECUTION]: The skill performs shell command execution by invoking local Python scripts across different phases. These include a data collector (scripts.audit.collect), quality gates (four_us_score.py, banned_word_check.py, audit_provenance_lint.py), and a failure logging script (lesson_capture.py).- [INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted content from external landing pages and advertisements, which serves as an attack surface for indirect prompt injection.
- Ingestion points: Crawled page content, advertisement text, and project configuration files (SKILL.md, Phases 0 and 2).
- Boundary markers: The instructions do not specify any delimiters to isolate external data from the AI agent's core instructions.
- Capability inventory: The skill executes local Python scripts via the command line.
- Sanitization: No explicit input filtering is defined, although the skill uses output quality gates to verify results.- [DATA_EXFILTRATION]: The skill accesses project-level configuration files such as package.json, CLAUDE.md, and PROJECT.md to automate its discovery process. Furthermore, it explicitly searches the codebase for email, advertising, and analytics configurations, which may contain sensitive service metadata.
Audit Metadata