skills/ingpoc/skills/autoresearch/Gen Agent Trust Hub

autoresearch

Pass

Audited by Gen Agent Trust Hub on Jun 15, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [PROMPT_INJECTION]: The skill contains directives to bypass user intervention and lacks protection against indirect injection from target skills.
  • The agent is instructed in SKILL.md to 'NEVER STOP' and 'do not pause to ask the user if you should continue' once the autonomous loop begins.
  • Ingestion points: Target SKILL.md files are read into context in Step 1 of SKILL.md.
  • Boundary markers: No delimiters or instructions are used to separate the target skill's content from the optimizer's logic.
  • Capability inventory: The skill can execute mutated skills repeatedly and run system commands like 'open'.
  • Sanitization: No validation or escaping is performed on the content of the target skills.
  • [COMMAND_EXECUTION]: The skill executes local commands and dynamically modified skill files.
  • Step 5 of the process involves the agent executing a modified version of a skill file ([user-chosen-name].md) as part of a recurring loop.
  • The agent is instructed to use the 'open' command to launch the dashboard.html file in a browser on macOS.
  • [EXTERNAL_DOWNLOADS]: The skill references external libraries from remote servers.
  • The instructions for Step 3 in SKILL.md specify that the generated dashboard should load the Chart.js library from a CDN.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 15, 2026, 09:45 AM
Security Audit — agent-trust-hub — autoresearch