content-hash-cache-pattern

Pass

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [SAFE]: The implementation provides a standard and secure approach to file-based caching using SHA-256 content hashes. The code utilizes established best practices for efficiency and data integrity.\n- [SAFE]: Data persistence is handled using the standard json module, avoiding unsafe alternatives like pickle. The implementation uses chunked reading to prevent memory exhaustion when processing large files.\n- [INDIRECT_PROMPT_INJECTION]: The skill provides a template for processing external files, which is an inherent part of the primary caching function. \n
  • Ingestion points: External file contents are read and hashed via compute_file_hash in SKILL.md.\n
  • Boundary markers: Not present in the generic template code.\n
  • Capability inventory: Filesystem read/write access via pathlib and open methods.\n
  • Sanitization: Content validation is outside the scope of this caching pattern.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 14, 2026, 11:08 PM
Security Audit — agent-trust-hub — content-hash-cache-pattern