github-agent-bridge

---

name: github-agent-bridge description: Operate and maintain the github-agent-bridge deployment: inspect queue health, reader/executor status, policy and routing, shadow/canary rollout, monitor alerts, retries, stale locks, autoupdate planning, feedback rules, and dashboard/systemd services. Use when working on bridge operations, production maintenance, rollout, troubleshooting, or safe release updates.

GitHub Agent Bridge

Use this skill when operating github-agent-bridge, the GitHub notification to OpenClaw agent bridge.

Operator Stance

• Prefer the short CLI name: gab.
• Default to safe inspection, shadow, dry-run, or --no-systemd commands unless the user explicitly asks for live production action.
• Do not pass --mark-seen while testing or shadowing IMAP; it mutates the mailbox.
• Keep secrets in the private environment file, never inline in commands or committed files.
• Treat enabledRepos as the live canary allowlist when it is present.
• Before unlocking or retrying jobs, inspect enough context to avoid duplicating active agent work.

Runtime Paths

Typical deployment paths:

DB=~/.local/state/github-agent-bridge/bridge.sqlite3
POLICY=~/.config/github-agent-bridge/policy.json
ENV=~/.config/github-agent-bridge/env

Repository docs to load as needed:

• Install or reconfigure a deployment: ../../../docs/installation.md
• Operate production, monitor, update, dashboard services: ../../../docs/operations.md
• Configure trust, actions, routing, roles, and models: ../../../docs/policy-reference.md
• Roll out safely through replay, shadow, dry-run, and canary: ../../../docs/shadow-canary.md
• Diagnose known failures: ../../../docs/failure-modes.md
• Understand architecture and invariants: ../../../docs/architecture.md
• Develop or test bridge changes: ../../../docs/development.md
• Release and changelog behavior: ../../../docs/releases.md

First Checks

Start with read-only state:

gab --db "$DB" status
gab --db "$DB" jobs --limit 20
gab --db "$DB" monitor --no-systemd
gab --db "$DB" monitor --json

For systemd deployments:

systemctl --user status github-agent-bridge.service
systemctl --user status github-agent-bridge-reader.timer
systemctl --user status github-agent-bridge-monitor.timer
journalctl --user -u github-agent-bridge.service -n 120 --no-pager

Safe Validation

Use an isolated DB for local or canary checks:

gab --db /tmp/github-agent-bridge-shadow.sqlite3 init-db
gab --db /tmp/github-agent-bridge-shadow.sqlite3 --policy "$POLICY" run --mode shadow --once
gab --db /tmp/github-agent-bridge-shadow.sqlite3 jobs --limit 50

Read IMAP without mailbox mutation:

set -a
. "$ENV"
set +a

gab --db /tmp/github-agent-bridge-shadow.sqlite3 \
  --policy "$POLICY" \
  read-imap-once \
  --email "$GITHUB_AGENT_BRIDGE_EMAIL" \
  --password "$GITHUB_AGENT_BRIDGE_PASSWORD" \
  --mailbox "$GITHUB_AGENT_BRIDGE_MAILBOX"

Only add --mark-seen when the bridge should consume those GitHub notifications.

Live Operation

Executor pool:

gab --db "$DB" --policy "$POLICY" \
  run --mode live --workers 4 --review-timeout 900 --work-timeout 3600

One-shot reader job:

gab --db "$DB" --policy "$POLICY" \
  read-imap-once \
  --email "$GITHUB_AGENT_BRIDGE_EMAIL" \
  --password "$GITHUB_AGENT_BRIDGE_PASSWORD" \
  --mailbox "$GITHUB_AGENT_BRIDGE_MAILBOX" \
  --mark-seen

The packaged github-agent-bridge-reader-run wrapper is preferred for systemd reader timers because it reads the private env file and handles IMAP mailbox quoting.

Queue Recovery

Inspect before acting:

gab --db "$DB" jobs --status blocked --limit 20
gab --db "$DB" jobs --status running --limit 20
gab --db "$DB" monitor --json

Typical recovery commands:

gab --db "$DB" retry <job-id>
gab --db "$DB" dismiss <job-id>
gab --db "$DB" unlock-stale --older-than 3600
gab --db "$DB" unlock-stale --job-id <job-id>

Use unlock-stale only after confirming no live OpenClaw process is still working on that job.

Updates

Plan before applying:

gab --db "$DB" update --repo-dir /path/to/github-agent-bridge --json
gab --db "$DB" update --repo-dir /path/to/github-agent-bridge --record --json

Apply only the safe immediate subset:

gab --db "$DB" update --repo-dir /path/to/github-agent-bridge --record --apply --json

Complete a deferred executor reload after the active queue drains:

gab --db "$DB" update --complete-pending --json

For non-standard services or install sources, prefer environment variables in the private env file or documented CLI flags over hardcoded command edits.

Policy And Feedback

Validate policy shape before live runs:

python3 -m json.tool "$POLICY" >/dev/null
gab --db "$DB" --policy "$POLICY" monitor --no-systemd

Inspect learned feedback rules that affect agent prompts:

gab --db "$DB" feedback-rules
gab --db "$DB" rules pilipilisbot/github-agent-bridge
gab --db "$DB" feedback-events --limit 20
gab --db "$DB" feedback-proposals --limit 20

Development Checks

When changing the bridge repository:

python3 -m venv .venv
. .venv/bin/activate
python -m pip install -e '.[test]'
pytest -q

For CLI behavior changes, also exercise the relevant command against a temporary DB.