cy-workflow-memory

Pass

Audited by Gen Agent Trust Hub on Apr 15, 2026

Risk Level: SAFEPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
  • [PROMPT_INJECTION]: The skill is vulnerable to Indirect Prompt Injection because it explicitly directs the agent to 'Treat these files as mandatory context for the run, not optional notes' (SKILL.md). This creates a risk where malicious instructions embedded in memory files could be followed by the agent.\n
  • Ingestion points: Workflow instructions in SKILL.md require reading shared and task-specific memory files.\n
  • Boundary markers: Absent. The skill provides no instructions to use delimiters or to ignore embedded commands within the memory content.\n
  • Capability inventory: The skill performs file system read and write operations.\n
  • Sanitization: Absent. No validation, filtering, or escaping of memory file content is performed.\n- [DATA_EXFILTRATION]: The skill accepts file paths as input ('Workflow memory directory path', 'Shared workflow memory file path') without performing validation or restricting access to specific directories. This could allow a caller to point the agent to sensitive local files (e.g., SSH keys, environment variables) to be read into the agent's context under the guise of workflow memory.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 15, 2026, 12:15 AM