super-human
Pass
Audited by Gen Agent Trust Hub on Mar 23, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill performs automated codebase convention detection by scanning the file system for various project configuration files, lock files (e.g., package-lock.json, Cargo.lock, go.sum), and directory patterns.
- [COMMAND_EXECUTION]: Upon activation, the skill executes a shell command using
lsto check for the presence of recommended companion skills in both home directory paths (~/.claude/skills/, ~/.agent/skills/, etc.) and local project directories. - [COMMAND_EXECUTION]: The skill instructs the agent to run project-specific development scripts for testing (e.g., npm test, pytest), linting, and building as part of the verification and convergence phases.
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests untrusted data from the local repository, including documentation files (docs/*.md) and source code, which are processed during the research and planning phases without explicit sanitization markers.
- Ingestion points: SKILL.md (Codebase Convention Detection), Phase 3 (DISCOVER) reading files and documentation.
- Boundary markers: Absent from prompts when reading external content.
- Capability inventory: File system access, shell command execution (npm/make/pytest), and board persistence writing.
- Sanitization: No explicit escaping or sanitization of ingested file content before interpolation into task prompts.
Audit Metadata