capability-evolver

Warn

Audited by Gen Agent Trust Hub on Mar 17, 2026

Risk Level: MEDIUMREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill facilitates autonomous code generation and 'self-evolution,' including an optional configuration (EVOLVE_ALLOW_SELF_MODIFY) that allows the AI to modify its own source code at runtime.
  • [COMMAND_EXECUTION]: The skill is designed to be executed via Node.js (node index.js) and suggests the use of npx capability-evolver, which executes code directly from the npm registry.
  • [PROMPT_INJECTION]: The instructions include identity-shaping directives such as 'You are a Recursive Self-Improving System' and 'Mutation Directive,' which are designed to override the agent's default behavior and safety constraints.
  • [PROMPT_INJECTION]: An indirect prompt injection surface is created by the 'Auto-Log Analysis' feature, which processes untrusted interaction history and logs to determine how the agent should evolve or patch its own code.
  • [DATA_EXFILTRATION]: To function, the skill requires read access to sensitive agent memory and history files, exposing all past interactions to the autonomous evolution engine.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 17, 2026, 07:18 PM