compound-learnings

Pass

Audited by Gen Agent Trust Hub on Jun 22, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill has an indirect prompt injection surface because it reads untrusted content from local files (git diffs, design documents, and plans) to update project configuration files like CLAUDE.md. The risk is mitigated by explicit instructions to the agent to treat file content as data only and ignore embedded instructions. Furthermore, Phase 2c implements strict sanitization and redacts known secret patterns (e.g., sk-, AKIA, ghp_).
  • [COMMAND_EXECUTION]: The skill uses local shell commands (git log, git diff, git symbolic-ref) to identify changes in the repository. These operations are limited to metadata extraction and are used for the skill's primary purpose of analyzing development work.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 22, 2026, 05:02 PM
Security Audit — agent-trust-hub — compound-learnings