agents-memory

Pass

Audited by Gen Agent Trust Hub on Oct 17, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill is a collection of best practices and defensive auditing scripts. No malicious behavior, data exfiltration, or unauthorized command execution patterns were detected.
  • [PROMPT_INJECTION]: Instructions regarding agent behavior focus on defensive minimalism and completeness maximalism as legitimate operating modes. There are no attempts to bypass safety filters or override system instructions maliciously.
  • [DATA_EXPOSURE]: The skill includes scripts/lint_claude_memory.sh, which actively scans for hardcoded secrets in instruction files to prevent accidental exposure. It also warns against including memory files in production build artifacts.
  • [COMMAND_EXECUTION]: The provided bash scripts are utility tools for local auditing. They use standard Unix utilities and do not perform any untrusted remote executions or network operations.
  • [INDIRECT_PROMPT_INJECTION]: The documentation identifies and warns against indirect prompt injection risks, specifically through vendored dependencies or unreviewed PRs. It provides clear mitigation strategies for these risks.
  • [OBFUSCATION]: All scripts and documentation files are written in clear, plain text. No obfuscation techniques were found.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 17, 2026, 08:31 AM