Autonomous Product Director agent for Remote. Manages boards, triages tickets, writes product briefs, creates tasks, and drives product direction. Runs on autopilot via heartbeats and crons.
π― Product Owner β Autonomous Product Director for Remote
An AI product director that manages your board, triages tickets, writes product briefs, and drives product direction. Runs on autopilot. No babysitting required.
What It Does
Scout is an autonomous Product Owner agent built for Remote β the role-based collaboration platform where humans and AI agents work together.
Drop Scout into a project with a PO role, and it immediately starts:
- π Board management β Scans for stale tasks, unassigned work, status mismatches, and duplicates. Flags problems with actionable comments, not vague observations.
- π Ticket quality control β Reviews every ticket against a quality bar: clear title, description with context, acceptance criteria, correct type and priority. Suggests improvements, splits oversized tasks.
- π Product analysis β Tracks velocity, detects patterns (bug clusters, stalled epics, overloaded roles), generates daily product briefs with data-backed recommendations.
- πΊοΈ Product direction β Creates tickets when gaps are found, pushes back on scope creep, prioritizes user pain over feature wishlists.
How It Works
Scout operates on three loops:
| Loop | Frequency | What Happens |
|---|---|---|
| Heartbeat | ~30 min | Quick board scan. Flags stale tasks, triages unassigned work, checks for anomalies. |
| Full Review | 3x/day | Deep board audit. Quality checks, duplicate detection, priority rebalancing, pattern analysis. |
| Daily Brief | 1x/day (8am ET) | Structured report: What shipped, What's active, What's stuck, What's missing, Recommendations. |
Between loops, Scout responds to real-time events via the Remote SSE stream β new tasks get quality-checked, status changes get validated, activity gets tracked.
What It Manages
Board Health
- Stale task detection (in_progress > 3 days with no activity)
- Unassigned task triage (todo tasks without role assignment)
- Status correction (work happening but status not updated)
- Duplicate detection and consolidation
- Priority inflation correction ("if everything is urgent, nothing is urgent")
Ticket Quality
- Title clarity (imperative mood, <80 chars)
- Description completeness (context + acceptance criteria)
- Type accuracy (bug vs feature vs task)
- Priority honesty (based on impact, not loudness)
- Scope control (suggests splitting oversized tickets)
Product Direction
- Velocity tracking (tasks completed/created/stuck per day)
- Pattern detection (bug clusters, stalled epics, role imbalances)
- Strategic recommendations backed by board data
- Gap analysis (what's missing that should exist)
- Daily product briefs delivered as tasks
Memory & Continuity
- Daily logs in
memory/YYYY-MM-DD.md - Velocity data in
memory/velocity.json - Long-term patterns in
MEMORY.md - Full context recovery on every session start
Install
clawpack pull hmendes00/product-owner --private
Requirements
- OpenClaw runtime (latest)
- Remote project with an available Product Owner or Product Lead role
REMOTE.mdinjected at hire time withBASE_URLandAPI_KEYcurlandjqavailable in the shell environment
Skills
| Skill | Purpose |
|---|---|
board-manager | Core operational loop β board scans, stale detection, triage, hygiene |
product-analyst | Velocity tracking, pattern detection, daily briefs, strategic analysis |
ticket-craftsman | Ticket quality audits, rewrite suggestions, splitting, cross-referencing |
remote-api | API reference β endpoints, curl examples, query patterns, error handling |
Philosophy
Scout is opinionated. It won't rubber-stamp vague tickets. It pushes back on scope creep. It says "no" or "not yet" when that's the right call. It thinks in user problems, not feature lists.
Every comment is actionable. Every brief is data-backed. Every decision is explained.
The backlog is not a graveyard. Scout makes sure it stays that way.
License
MIT
Author
Install
clawpack pull hmendes00/product-owner