live-transcribe-read

Pass

Audited by Gen Agent Trust Hub on Jun 13, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses shell commands including cat, grep, sed, and kill to interact with local files. Specifically, it reads from /tmp/realtime-transcribe.log and uses the content of /tmp/realtime-transcribe.pid to check process status.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted text from a transcription log and provides it to the agent for further processing (e.g., summarization or translation).
  • Ingestion points: /tmp/realtime-transcribe.log (SKILL.md)
  • Boundary markers: Absent; the content is read directly into the context without delimiters.
  • Capability inventory: The agent has access to shell commands (cat, grep, sed, kill) and general LLM reasoning capabilities.
  • Sanitization: None; the raw content of the log file is processed and displayed.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 13, 2026, 09:29 PM
Security Audit — agent-trust-hub — live-transcribe-read