hdresearch

Use when doing code work that must be integration-friendly: branches, commits, tests, PR-ready receipts, owned paths, and parent-friendly handoff.

Use when implementing and exposing a product/application service or UI that should run outside Reef root. Choose child/resource VM placement, stand up the app, and report how to reach it.

Use when handling parent commands such as steer, abort, pause, or resume, or when deciding how urgently to react to inbox messages from above.

Use when agents need to coordinate through reef store, peer signals, barriers, rendezvous, or child/peer communication without breaking authority boundaries.

Create a new service module for reef. Use when adding a new capability to the server — a new store, API routes, LLM tools, behaviors, or dashboard widget.

Use when a task has multiple independent subsystems, needs recursive delegation, or requires a parent to split implementation and then integrate the results.

Use when inspecting the live fleet, differentiating active vs historical lineage, tracing ancestry, checking subtree state, or doing post-mortem investigation.

Use when working with GitHub repos, branch setup, auth tokens, PR flow, or repo preparation inside Reef.

Always register VM snapshots in the commit ledger. Use when committing/snapshotting VMs, before destroying VMs, during deploys, or when creating golden images. Ensures no snapshot goes untracked.

Generate a magic login link for the agent-services dashboard UI. Use when the user wants to access the dashboard, needs a login link, says "magic link," or asks for UI access.

Shared task board for coordinating work across agents. Use when creating, assigning, tracking, or querying tasks in a swarm.

Deploy a branch to a preview VM for zero-risk change review. Snapshot infra, clone it, deploy the branch on the clone, share the link. Production untouched. Use for ALL UI changes, feature demos, and PR reviews.

Commit ledger for tracking VM snapshots. Use when recording, querying, or managing Vers VM commit history — golden images, infra snapshots, rollback points.

Append-only work log for agent sessions. Use when recording what happened, what was decided, and what's next — the log is for the next session's orchestrator.

Deploy agent-services to the infra VM. Use when shipping new code, restarting services, or rolling back a broken deploy. Covers pre-deploy snapshot, pull/build/restart, smoke tests, and rollback.

Create, view, and share agent reports. Use when publishing structured findings, generating share links for stakeholders, or tracking who accessed a shared report.

Bootstrap a self-improving agent fleet on Vers from scratch. Walks through deploying agent-services, building a golden image, testing the loop, and running first tasks. Use when onboarding someone new to the Vers agent fleet workflow.

Bootstrap a Vers VM into a golden image with punkin-pi, Node.js, dev tools, and swarm extensions installed. Creates a committed snapshot that can be branched for self-organizing agent swarms.

How Vers VM networking works — public URLs, port routing, TLS proxy, IPv6, and running services on VMs.

Contribute a bug fix or improvement to pi-v. Use when you've found and fixed an issue in the pi-v package.

Orchestrate agent swarms across Vers VMs - spawn parallel agents on branched VMs for distributed work, parallel exploration, or multi-path problem solving. Use when you need to run multiple agents simultaneously on isolated environments.

Manage Vers VMs (vers.sh) - create, branch, commit, restore, pause/resume Firecracker VMs. Use when working with Vers platform, VM orchestration, or when you need isolated execution environments.

Deep investigation of Vers platform issues (API, orchestrator, agent, docs). Use when encountering any Vers platform problem that needs thorough debugging and issue reporting.

Guidelines for Vers platform development. Agent Experience (AX) is a product metric - investigate ALL Vers issues (API, orchestrator, agent, docs) rather than silently working around them. Covers issue triage, reporting templates, and communication patterns.

Use Vers as an agent-subdelegation substrate: one local supervisor farms work to planner, codebase-minimap, reviewer, researcher, or execution subagents running in Vers VMs. Trigger on subagent, delegate, farm out work, remote agents, supervisor/worker, plan+workers, parallel agent decomposition, long-running autonomous task, or agent swarm on Vers.

Load when local compute is the wrong place for a task. Examples: persistent services, one-off heavy jobs, recurring builds, side experiments, matrices of configurations or builds run in parallel, middleware test boxes, preserving battery when running on portable power, building your own Beowulf cluster, or implementing the world's most expensive Prolog logic search possible. Loading does not allocate Virtual Machines (VMs). Run the reach gate, keep small or sensitive work local, tell the user before creating any VM, and preserve work product before ending any VM.

Orchestrate multiple AI agents across Vers VMs for parallel task execution

Write system prompts, tool docs, and agent definitions. Combines research-backed prompt engineering (+15-30% measured improvements) with project XML conventions. Covers tag hierarchy, structural templates, high-impact interventions, anti-patterns.