Developer Assistant

The On-Call Engineering Partner That Never Sleeps

Code review backlogs cost velocity. Untriaged bugs cost morale. CI failures that sit for hours cost trust. The Developer Assistant is an AI agent blueprint that runs continuously alongside your engineering workflow — reviewing PRs, triaging issues, and alerting on build failures — so your team ships faster with fewer manual coordination tasks.

This blueprint provides a pre-configured OpenClaw agent with deep GitHub and CI integration, calibrated specifically for technical work.

What the Agent Does

PR Review (every 30 minutes): The agent scans all open PRs in your configured repos that haven't been reviewed yet. For each one, it reads the full diff and checks for: actual bugs (logic errors, race conditions, null handling), security vulnerabilities (injection risks, exposed secrets, auth bypasses), missing error handling, performance issues, and gaps in test coverage. It posts inline GitHub comments with specific line references and a plain-English explanation of each issue. It then posts a summary comment with an overall assessment and a verdict label (approved / needs-changes / needs-tests).

The review philosophy, encoded in SOUL.md, is deliberately focused: correctness and security are always reviewed; style is ignored unless egregious. The agent won't block a PR over a naming preference.

CI Monitoring (every 5 minutes): Checks build status on main and develop branches. On failure, it sends an immediate Telegram alert with: which repo, which branch, which job failed, and the specific failed step. When CI recovers, it sends a confirmation. You'll know about a broken build within 5 minutes of it happening, not at the next standup.

Issue Triage (every hour): Monitors for new unlabeled GitHub issues. Classifies each as bug/feature/question/duplicate, assigns a severity (P0–P3), selects appropriate labels, suggests an assignee based on ownership, and posts a triage comment. P0 and P1 issues fire a Telegram alert immediately.

Daily Standup (9 AM): Compiles a summary of yesterday's merged PRs, PRs awaiting review today, open critical issues, and overall CI health — delivered to your team Telegram channel.

Who This Is For

Engineering teams of 2–15 who spend meaningful time each week on manual code review scheduling, issue backlog grooming, and checking build status. Also valuable for solo developers who want automated code review as a second pass before merging.

What's Included

Setup Steps

1. Copy the blueprint into ~/.openclaw/workspace/
2. Edit USER.md — add your GitHub repos, team handles, and ownership areas (who owns auth, API, frontend, etc.)
3. Add GitHub token to ~/.openclaw/secrets/github.json with repo read/write and PR comment permissions
4. Configure Telegram alerts — set P0 alert channel and standup digest channel
5. Set code standards in USER.md — the agent uses these when reviewing PRs
6. Start the agent — openclaw start

What "Senior-Level" Review Actually Means

The agent is configured to catch the things senior engineers catch in code review: missing null checks on async calls, SQL injection via string concatenation, unbounded loops with external data, race conditions in concurrent operations, hardcoded secrets or credentials. It cites the specific line, explains the risk, and suggests a fix — not just flags the problem.

Why This Beats Manual Process

A human code reviewer takes 20–60 minutes to review a substantial PR thoroughly. With this agent, every PR gets a first-pass review within 30 minutes, 24/7, including at 2 AM when someone pushes a hotfix. Human reviewers then focus on architecture decisions, business logic, and the edge cases the agent flagged — not on scanning for obvious issues.

Result: fewer bugs shipped, faster PR turnaround, and a CI health that's monitored in real time instead of discovered at standup.