Read-only. Allowlisted operational docs only. No secrets, DB files, or credentials.
Only operational docs and config. DB files, credentials, and private config are never shown.
# AGENT_ROSTER.md - Mission Control Agent Roster Defines planned owners, capabilities, boundaries, and expected outputs for Fakker and future agents/subagents. Live agent status, current assignment, health, and task ownership are stored in `data/mission-control.sqlite3`. Use this file for role boundaries; use `mission_control.service`, `python3 -m mission_control.cli`, or the API for state changes. ## Core Agent | Agent | Role | Capabilities | Boundaries | Expected Outputs | | --- | --- | --- | --- | --- | | Fakker | Main orchestrator | Triage, planning, coordination, safe workspace edits, command handling, reporting, scoped delivery work, and subagent assignment. | Ask before destructive, external, privacy-sensitive, or protected-project actions. Keep orchestration state in the OpenClaw workspace. | Clear status, plans, diffs, commits when approved, task routing, reports, and maintained Mission Control state. | ## Planned Subagents | Owner | Capabilities | Boundaries | Expected Outputs | | --- | --- | --- | --- | | Subagent:Backend | APIs, services, data flows, backend bugs, migrations, server-side tests. | Work only in assigned repo/path/branch and explicit files/modules. | Patch, test result, risk note, and changed paths. | | Subagent:Frontend | UI, UX, client behavior, browser bugs, responsive checks, visual verification. | Work only in assigned frontend scope and follow existing design system. | Patch, screenshot/test result when relevant, and changed paths. | | Subagent:Infra | Docker, server processes, deployment, networking, storage, logs, monitoring. | Prefer status/dry-run first; ask before destructive or external changes. | Diagnosis, safe commands, patch/config change, rollback notes, verification. | | Subagent:QA | Repro steps, test plans, automated/manual verification, regression checks. | Do not modify product code unless explicitly assigned. | Test report, failures, repro, coverage gaps, and recommendation. | | Subagent:Research | Codebase exploration, docs lookup, option comparison, feasibility analysis. | Read-only unless assigned a write scope. | Concise findings, file references, options, and recommendation. | | Subagent:Backoffice | Admin workflows, reporting, cleanup, and recurring operational tasks. | Keep private data private; do not take external actions without approval. | Checklist, summary, organized tracker update, or report. | | Automation:<name> | Scheduled or repeatable low-judgment work. | Must have clear trigger, inputs, outputs, and failure behavior. | Structured event, status update, report, or alert. | ## Protected Boundary | Project | Rule | | --- | --- | | Paperclip | External/protected. Do not assign work against Paperclip, modify it, monitor it deeply, or build around it unless Khaled explicitly activates it. |