lyric-diagnostic

Warn

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: MEDIUMDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [DATA_EXFILTRATION] (MEDIUM): Arbitrary File Read and Path Traversal vulnerability detected in cliche-check.ts and meter-check.ts. Both scripts use readFileSync on an unvalidated file path provided via the --file command-line argument (e.g., readFileSync(args[1], "utf-8")). This allows the agent to be coerced into reading and displaying the contents of sensitive files (e.g., .env, SSH keys, or system configurations) if an attacker provides a malicious path.
  • [PROMPT_INJECTION] (LOW): Vulnerability to Indirect Prompt Injection. The scripts ingest untrusted lyric data and output it back to the agent without sanitization or boundary markers. Ingestion points: Command-line arguments and file input in both scripts. Boundary markers: Absent. Capability inventory: Local file system read. Sanitization: Absent. Malicious instructions hidden within lyrics could influence the agent's behavior when it processes the tool's output.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 17, 2026, 05:32 PM