God's eye view of your dev repos. Multi-project tracking across GitHub/Azure DevOps. AI learns from your commits to upgrade your agents.md.

**Quick judgment** This skill is a strong fit for developers or teams managing multiple GitHub-based projects who want both a portfolio-level status view and structured feedback on how to improve their AI agent instructions (e.g., `agents.md`) based on real work patterns. It is opinionated toward CLI-centric, local workflows and integrates naturally into OpenClaw agents that operate over your repos. **What it does well** - **Multi-project oversight:** `god status` provides a bird’s-eye view across all configured repos and can drill into a single project for recent commits, PRs, and issues. - **Efficient, local data model:** Incremental sync via `god sync` plus a local SQLite cache (`~/.god-mode/cache.db`) gives reasonable performance for ongoing use without re-pulling full histories. - **Agent instruction coaching:** `god agents analyze <project>` compares `agents.md`-style files against commit history to find gaps between declared agent behavior and actual development patterns, then suggests improvements. - **Local-first and auth delegation:** It stores data only under `~/.god-mode` and defers authentication to existing CLIs (e.g., `gh auth login`), avoiding direct token handling. **Limitations and constraints** - **Provider support:** GitHub via `gh` is the only fully supported provider; Azure DevOps and GitLab are listed as “coming”, so current multi-host coverage is incomplete. - **Environment assumptions:** Requires `gh`, `sqlite3`, and `jq` binaries plus a local clone path where relevant; this is ideal for developer workstations but less plug-and-play on constrained environments or managed runners. - **Feature maturity:** Some capabilities (e.g., `god agents generate <project>`) are explicitly “Coming Soon”, so it’s focused today on analysis, not bootstrapping. **Risks and safety considerations** - **Exposure of sensitive repo data:** By design, the tool aggregates commit, PR, and issue data across all configured projects. If agents are allowed to call it freely, they can surface private or sensitive repository information in their responses. Use it only in trusted contexts and restrict which agents can access it for proprietary codebases. - **Scope of analysis files:** The default `analysis.agentFiles` includes common instruction files like `agents.md`, `CLAUDE.md`, and `.github/copilot-instructions.md`. Ensure these files do not contain secrets, and be aware that suggested changes may cause agents to rely more heavily on patterns in your private repos. - **Reliance on external CLIs:** Misconfiguration or changes in `gh`/`az`/`glab` behavior can break sync or authentication, which may cause silent data staleness if agents are not checking for sync freshness. **Recommended scenarios** - You actively use **GitHub** and want a **daily or weekly portfolio dashboard** (`god status`) for multiple repos, including stale PR detection and quiet-project identification. - You maintain **agents or copilot-style instructions per repo** and want a data-driven way to refine them based on actual commit/PR patterns (`god agents analyze`). - You’re building **OpenClaw workflows** where one or more agents regularly: - Run `god status` to generate heartbeat/standup summaries (e.g., for `HEARTBEAT.md`). - Run `god agents analyze` on high-priority projects on a schedule and notify when misalignment is detected. If your work is GitHub-centric and you’re comfortable with CLI tooling, this skill is a solid, relatively safe and performant choice for multi-project oversight and agent-instruction improvement; just be deliberate about which agents can see what it surfaces from private repositories.