hindsight-architect
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFECREDENTIALS_UNSAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [CREDENTIALS_UNSAFE]: The preamble script accesses sensitive local files, specifically reading .env and ~/.hindsight/config, to identify API keys and service URLs for environment detection.
- [DYNAMIC_EXECUTION]: The skill preamble dynamically locates and executes a local binary (hindsight-update-check) from computed paths within the user's home directory across multiple potential agent skill storage locations (~/.claude/skills, etc.).
- [COMMAND_EXECUTION]: The skill instructs the agent to execute a bash script (the 'Preamble') at the start of the session, which performs file system traversal and binary execution.
- [INDIRECT_PROMPT_INJECTION]: The skill processes project-specific data (README, manifests, and source code) to generate recommendations, which introduces a surface for instructions embedded in those files to influence agent behavior.
Audit Metadata