sdd-propose

Pass

Audited by Gen Agent Trust Hub on Apr 8, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection by ingesting data from multiple untrusted sources without sufficient safeguards.\n
  • Ingestion points: The skill reads all markdown files within the ai-context/features/ directory (Step 0b) and architectural documentation from ai-context/architecture.md (Step 1). It also retrieves prior data from the engram memory service via mem_search and mem_get_observation (Steps 1, 3a, 5).\n
  • Boundary markers: No delimiters or instruction-ignore warnings are specified to separate external data from the agent's core instructions during the proposal generation process.\n
  • Capability inventory: The skill has the capability to write to the system's memory service using the mem_save tool (Step 3b).\n
  • Sanitization: There is no evidence of validation, escaping, or filtering applied to the content retrieved from local files or memory before it is processed by the model.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 8, 2026, 05:45 AM