kai-local-audit
Pass
Audited by Gen Agent Trust Hub on Oct 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests and processes untrusted data from external websites and third-party AI recommendations, creating a vulnerability surface for indirect prompt injection.
- Ingestion points: The skill reads external business websites (Phase 0), performs site crawls (Phase 2), and aggregates recommendations from external AI assistants (Phase 2).
- Boundary markers: Instructions explicitly require the agent to verify competitor positioning claims and AI-suggested business information against the target business's own site before reporting.
- Capability inventory: The agent can execute local shell commands via internal Python scripts and write structured data to the local workspace.
- Sanitization: While the skill mandates logical verification of claims, it lacks explicit instructions for technical sanitization or escaping of scraped HTML/text before it is interpolated into the final HTML report.
- [COMMAND_EXECUTION]: The skill requires the execution of multiple internal Python modules for SEO data retrieval, Lighthouse audits, and system checks. These commands are part of the core auditing functionality.
- [EXTERNAL_DOWNLOADS]: The skill references a external repository for its full data harness. This repository is owned by the skill's author and represents a vendor-owned resource.
- Evidence: The skill notes that located metrics require the harness found at github.com/cgallic/kai-cmo-harness.
Audit Metadata