cohere-data-handling

Pass

Audited by Gen Agent Trust Hub on Sep 11, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill establishes patterns for processing untrusted user data that is eventually passed to external Cohere API endpoints. This creates a potential surface where malicious instructions embedded in the data could influence the agent's behavior.
  • Ingestion points: The functions safeCohereChat, safeEmbed, and safeClassify in SKILL.md ingest arbitrary text inputs from the user.
  • Boundary markers: The provided code examples do not include explicit prompt boundary markers or directives to ignore instructions within the API payload.
  • Capability inventory: The skill leverages the cohere-ai SDK to interact with remote chat, embedding, and classification services.
  • Sanitization: The skill implements a redactPII function that replaces detected sensitive information with placeholders, focusing on PII protection but not specifically on instruction sanitization.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 11, 2026, 04:15 PM
Security Audit — agent-trust-hub — cohere-data-handling