transcript

Fail

Audited by Gen Agent Trust Hub on Apr 29, 2026

Risk Level: HIGHCOMMAND_EXECUTIONCREDENTIALS_UNSAFEPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The authentication script scripts/tapi-auth.js implements a persistence mechanism by modifying shell configuration and environment files across multiple platforms.
  • Evidence: The script automatically modifies .zshenv, .zprofile, .profile, .bashrc, .config/environment.d/transcript-api.conf, .config/fish/config.fish, and Windows PowerShell profiles to inject environment variables.
  • [CREDENTIALS_UNSAFE]: The skill stores sensitive API keys in plain text within numerous shell profiles and agent configuration files, increasing the potential exposure of the user's credentials.
  • Affected files include: ~/.openclaw/openclaw.json, ~/.clawdbot/moltbot.json, and various shell RC files.
  • [PROMPT_INJECTION]: The skill has a surface for indirect prompt injection as it processes external content (YouTube transcripts) with broad system capabilities.
  • Ingestion points: YouTube transcript text fetched via the transcriptapi.com API in SKILL.md.
  • Boundary markers: Absent; there are no instructions to the agent to distinguish between its commands and the transcript content.
  • Capability inventory: Capability to execute shell commands (curl) and perform file-system write operations through the bundled Node.js script.
  • Sanitization: Absent; the skill does not validate or filter the transcript content before it is processed by the agent.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Apr 29, 2026, 04:00 AM