omh-codegraph-refresh
Pass
Audited by Gen Agent Trust Hub on Nov 9, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is primarily instructional and defines a structured workflow for local repository analysis. It uses standard development tools (git, ripgrep, ast-grep) and vendor-specific CLI tools (omh) for indexing and summarizing code without performing suspicious network requests or executing untrusted remote code.
- [CREDENTIALS_UNSAFE]: The skill includes an explicit safety directive to protect sensitive information: 'Never expose secret values from codegraph inputs or config files; record redacted paths and warning categories only.' This proactive instruction reduces the risk of accidental credential exfiltration during codebase analysis.
- [INDIRECT_PROMPT_INJECTION]: As a tool designed to process and summarize entire codebases, the skill has an inherent surface for indirect prompt injection. However, it mitigates this risk by requiring 'observed command evidence' before reporting results and establishing a 'not-evidence boundary' to prevent the agent from overclaiming architecture verification based on potentially poisoned or unverified data.
- [COMMAND_EXECUTION]: The skill utilizes shell commands for repository status checks (e.g., git log, git status, git reflog). These are used strictly for local cache invalidation (staleness checks) and are performed within the expected scope of a codebase management utility.
Audit Metadata