anthropic-prompt-caching-and-context-runtime

Pass

Audited by Gen Agent Trust Hub on Jun 17, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill provides structured guidance for implementing Anthropic's Messages API features, specifically prompt caching and context management. It does not contain executable code or request dangerous permissions.
  • [DATA_EXFILTRATION]: The instructions include robust safeguards against data exposure, explicitly mandating the redaction of PII, secrets, and raw prompt content from telemetry and logs.
  • [CREDENTIALS_UNSAFE]: The skill follows security best practices by instructing the agent and developer to use environment-injected credentials and warns against committing API keys to source control.
  • [PROMPT_INJECTION]: No prompt injection or behavior override patterns were detected. The skill's instructions are focused on technical implementation details and conformance to architectural standards.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 17, 2026, 08:35 AM
Security Audit — agent-trust-hub — anthropic-prompt-caching-and-context-runtime