front-refactor
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-provided source files and possesses broad file modification capabilities, creating an attack surface for instructions disguised as code comments.
- Ingestion points: The skill reads the full content of target files (JS, TS, CSS, Vue, Svelte, Astro) as defined in
SKILL.mdStep 4. - Boundary markers: There are no explicit delimiters or instructions provided to the agent to treat the content of the target files strictly as data and to ignore any natural language directives found within them.
- Capability inventory: The skill is granted
Read,Glob,Edit, andWritepermissions in theSKILL.mdfrontmatter, allowing for significant filesystem impact if an injection succeeds. - Sanitization: The skill does not perform any pre-processing, validation, or escaping of the target file content before it is analyzed by the language model.
Audit Metadata