transcript-cleanup

Pass

Audited by Gen Agent Trust Hub on Jun 13, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill processes raw transcript content which could contain malicious instructions designed to hijack the agent's logic during the text cleaning phase.\n
  • Ingestion points: Reads all .txt files in the output/ directory (SKILL.md, Step 1).\n
  • Boundary markers: Extracts content under specific headers (Step 2) but does not instruct the agent to ignore instructions embedded within the transcription text.\n
  • Capability inventory: Access to Write tool and Bash tool for system operations (frontmatter).\n
  • Sanitization: Applies formatting and linguistic cleanup rules (Step 4) but lacks logic to detect or neutralize adversarial prompts.\n- [COMMAND_EXECUTION]: The skill constructs and executes Bash commands using strings derived from file metadata and external tool outputs.\n
  • Ingestion points: Filenames from the output/ directory are used to build find and mv commands (SKILL.md, Step 8).\n
  • Boundary markers: None; filenames are directly interpolated into shell command strings.\n
  • Capability inventory: Uses Bash tool to execute date, find, mv, and rm commands.\n
  • Sanitization: No sanitization is performed on the filenames. If a file in the output/ directory is maliciously named with shell metacharacters, it could lead to arbitrary command execution when the find command is invoked in Step 8.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 13, 2026, 02:32 AM
Security Audit — agent-trust-hub — transcript-cleanup