adopt
Pass
Audited by Gen Agent Trust Hub on Apr 14, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes shell commands (echo and cat) within a dynamic context injection block in SKILL.md to display the architecture documentation path and version information upon loading. These commands are used to provide local context to the agent and do not involve network exfiltration or sensitive file access.
- [PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it processes untrusted project data to perform audits and code modifications.
- Ingestion points: Reads files from the target project directory and architectural guidance from the local architecture folder.
- Boundary markers: Employs a structured checklist for the audit subagent but does not implement explicit 'ignore embedded instructions' markers for scanned project files.
- Capability inventory: Includes shell command execution at load time, extensive file writing in the tseng/ directory, and modification of project source files and CLAUDE.md.
- Sanitization: The skill does not perform sanitization or verification of the ingested project file content before processing it for decisions.
Audit Metadata