planning-with-files

Pass

Audited by Gen Agent Trust Hub on Mar 10, 2026

Risk Level: SAFENO_CODE
Full Analysis
  • [SAFE]: The skill consists of markdown instructions and templates for task planning. It contains no executable scripts, which significantly reduces the risk of technical exploits.\n- [DATA_EXFILTRATION]: There is no evidence of hardcoded credentials, access to sensitive file paths, or unauthorized network activity. The skill focuses on managing local files specific to the current task.\n- [PROMPT_INJECTION]: The instructions are designed to improve task performance and do not contain patterns intended to bypass safety filters or extract system prompts.\n- [PROMPT_INJECTION]: This skill possesses a vulnerability surface for indirect prompt injection because it ingests external data into findings.md and then reads it to inform decisions. Mandatory evidence chain:\n
  • Ingestion points: findings.md, references/findings.md\n
  • Boundary markers: Absent\n
  • Capability inventory: The skill instructions guide the use of the agent's native tools (e.g., browser, shell) for task execution.\n
  • Sanitization: Absent
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 10, 2026, 01:27 AM