autoresearch
Pass
Audited by Gen Agent Trust Hub on May 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill is designed to generate a shell script (
autoresearch.sh) and execute it usingbash -c. It also performs various filesystem and version control operations using shell commands such asgit,mkdir,cp, andmv. - [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection by ingesting untrusted data from the local repository.
- Ingestion points: Reads all project files in scope to "understand the workload deeply" (SKILL.md, Step 3) and parses output from the generated
autoresearch.shscript. - Boundary markers: No delimiters or instructions to ignore embedded commands are specified for the source files being read.
- Capability inventory: The agent has the ability to execute shell commands (
bash -c), write/modify files, and commit changes to the git repository. - Sanitization: There is no mention of escaping or validating the content of the project files before they are processed by the agent to influence the loop logic.
Audit Metadata