fix-on-my-computer

Warn

Audited by Gen Agent Trust Hub on Jul 24, 2026

Risk Level: MEDIUMCREDENTIALS_UNSAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [CREDENTIALS_UNSAFE]: The skill instructs the agent to access ~/.claude/.credentials.json during its troubleshooting phase. This file is a sensitive platform-specific path that may contain authentication tokens or API keys.
  • [COMMAND_EXECUTION]: Uses dynamic context injection in the SKILL.md file (!uname, !pwd, !bun) to execute shell commands immediately upon skill load, replacing markers with live system data.
  • [EXTERNAL_DOWNLOADS]: Automates the installation of Node.js dependencies using bun install and encourages the agent to dynamically add new packages (bun install <package>) to resolve missing dependency errors.
  • [PROMPT_INJECTION]: An indirect prompt injection surface is present because the skill processes external repository content (code and tests). A malicious repository could contain embedded instructions that exploit the agent's 'Fix Loop' capability to perform unauthorized file modifications or execute arbitrary commands via the Bash tool.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jul 24, 2026, 09:42 PM
Security Audit — agent-trust-hub — fix-on-my-computer