bootstrap
Pass
Audited by Gen Agent Trust Hub on Mar 25, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: Uses dynamic context injection (commands prefixed with '!') to probe the project environment, including checking for specific configuration files (e.g., 'package.json', 'Cargo.toml') and scanning for installed skills in the local directory.
- [EXTERNAL_DOWNLOADS]: Recommends the installation of additional skills and tools via the 'npx skills add' command, targeting resources within the author's own namespace ('jwilger/agent-skills').
- [COMMAND_EXECUTION]: Instructs the agent to automatically stage and commit generated instruction and configuration files to the local git repository to maintain workflow history.
- [PROMPT_INJECTION]: Implements a mechanism to aggregate instructions from external files ('agents-md-setup.md') found in other installed skills into the project's root 'AGENTS.md'.
- Ingestion points: Reads from 'skills/*/references/agents-md-setup.md'.
- Boundary markers: Uses '' delimiters to isolate generated content.
- Capability inventory: Includes shell command execution through dynamic injection and git version control operations.
- Sanitization: No explicit sanitization or validation of the aggregated markdown content is mentioned.
Audit Metadata