skill-developer
Pass
Audited by Gen Agent Trust Hub on Nov 5, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The documentation instructs the agent to execute shell commands, specifically using 'npx tsx' to run TypeScript files and 'chmod +x' to modify script permissions within the '.claude/hooks/' directory.- [INDIRECT_PROMPT_INJECTION]: The 'UserPromptSubmit' hook architecture creates a vulnerability surface by ingesting user prompts and injecting generated text directly into the agent's context.
- Ingestion points: User prompts are captured via stdin in '.claude/hooks/skill-activation-prompt.ts'.
- Boundary markers: Suggestions use ASCII borders, but there are no mandatory 'ignore embedded instructions' warnings for the injected content.
- Capability inventory: The skill guide describes subprocess execution via 'npx' and shell wrappers.
- Sanitization: There is no documentation regarding the sanitization or escaping of prompt data before it is processed by regex or injected into the context.- [DYNAMIC_EXECUTION]: The skill system relies on the dynamic execution of local TypeScript and shell scripts to enforce guardrails and trigger skills. This execution is triggered automatically during prompt submission or when using file-editing tools.
Audit Metadata