context-discovery

Pass

Audited by Gen Agent Trust Hub on Mar 5, 2026

Risk Level: SAFEPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it instructs the agent to read and follow instructions from local project files that may be attacker-controlled.
  • Ingestion points: Markdown files such as code-quality.md and security-patterns.md located within the resolved context_root.
  • Boundary markers: No delimiters or safety instructions are provided to distinguish ingested context from the agent's core instructions.
  • Capability inventory: The agent is directed to apply these discovered patterns to implementation tasks and pass them to subagents.
  • Sanitization: No validation or sanitization of the content within the discovered files is performed before adoption.
  • [DATA_EXFILTRATION]: The discovery protocol allows the context_root to be defined via a .oac.json file or discovered through a chain of glob patterns. This mechanism can be exploited for path traversal if a malicious configuration points the agent to sensitive directories (e.g., home directory paths) that contain a navigation.md file, potentially leading to unauthorized data access.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 5, 2026, 12:50 PM