write-hvac-authority-content
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill represents a surface for indirect prompt injection because it is designed to read, optimize, and write back content from local markdown files which are considered untrusted input.
- Ingestion points: The agent is instructed to read files from the
src/data/postandsrc/data/productdirectories using theread_filetool. - Boundary markers: The instructions do not specify any delimiters or warnings to ignore embedded instructions within the content files being processed.
- Capability inventory: The skill utilizes powerful filesystem tools including
write_file,edit_file, andcreate_file, allowing it to modify the repository based on processed data. - Sanitization: No sanitization or validation logic is defined to check the content of the files before the agent processes and acts upon the information.
- [COMMAND_EXECUTION]: The skill instructions include the execution of local shell commands (
npm run build,npm run check:astro) to verify content quality. While standard for Astro development, these commands execute in the project environment.
Audit Metadata