clawtrol

Imported from wolverin0/clawtrol (AGENTS.md). Install upstream with npx skills add wolverin0/clawtrol. Copyright stays with the author.

AGENTS.md - ClawTrol Operational Guide

This file is the operational source-of-truth for AI coding assistants working on ClawTrol.

1) Operational Context (Critical)

Use this map first. Most recent confusion came from editing the wrong folder.

Canonical vs mirror folders

UPDATED 2026-07-26 — Source edits remain local, while releases use the exact SHA-tagged image produced by hosted CI. The VM does not pull or build application source during a normal deployment.

Runtime endpoints

• ClawTrol app URL: http://192.168.100.186:4001
• Runtime: one Compose-managed web container; retired systemd web/worker units remain stopped.

Deploy

Use the /deploy-to-vm skill at .claude/skills/deploy-to-vm/SKILL.md. It deploys only an exact SHA image after hosted CI and restored-database migration gates pass, then requires /health to report the same SHA.

2) Mandatory Preflight Before Any Edit

All edits happen in the local clone at G:\_OneDrive\OneDrive\Desktop\Py Apps\clawtrol\. Run these checks locally in order before coding:

# 1) Confirm cwd is the local clawtrol clone
pwd && git rev-parse --show-toplevel
# Expected: .../clawtrol  (NOT clawtrol-workspace, NOT clawdeck_remote_work)

# 2) Confirm origin matches the canonical repo
git remote get-url origin
# Expected: https://github.com/wolverin0/clawtrol.git

# 3) Confirm branch + dirty state + sync with origin
git branch --show-current
git status --short
git fetch origin && git status -uno   # see if behind/ahead

# 4) Confirm target file exists locally
ls -la <target_path>

If pwd does not end in clawtrol OR origin is not wolverin0/clawtrol, stop and re-route. Do NOT SSH to the VM to edit. Pre-deploy verification of the VM state is handled by the /deploy-to-vm skill, not the edit preflight.

3) Documentation Source-of-Truth Policy

• Canonical roadmap lives in: docs/roadmaps/
• Execution evidence lives in: docs/reports/ and docs/artifacts/
• Session handoff context lives in: handoff.md

When roadmaps conflict:

1. Keep one canonical roadmap marked ACTIVE.
2. Mark older files as superseded, do not delete historical context.
3. Update checkbox status immediately after each completed step.

4) Project Overview

ClawTrol (formerly ClawDeck) is a Rails 8.1 mission control dashboard for AI agents.

Core capabilities:

• Task queue with board-based workflow
• Agent orchestration state (sessions, model routing, validation)
• Swarm idea launcher
• Factory loop automation
• ZeroBitch fleet controls
• Nightshift/cron orchestration

5) Architecture Snapshot

Stack

• Ruby 3.3.x / Rails 8.1
• PostgreSQL (primary/cache/queue/cable)
• Solid Queue, Solid Cache, Solid Cable
• Hotwire (Turbo + Stimulus) + Tailwind
• Puma + Nginx deployment

P0 Data Contract (Feb 2026)

tasks.description is the HUMAN BRIEF only. Agent output goes to TaskRun.

Key columns:

• tasks.description — Human task brief (never mutated by agents)
• tasks.original_description — Backup of original brief
• tasks.execution_prompt — Prompt for agent execution (was execution_plan)
• tasks.compiled_prompt — Pipeline-generated prompt from ERB templates
• task_runs.agent_output — Agent findings/output per run
• task_runs.prompt_used — Immutable snapshot of the prompt sent to agent
• task_runs.agent_activity_md — Markdown transcript summary
• task_runs.follow_up_prompt — Follow-up instructions for requeue

Prompt precedence chain: compiled_prompt || execution_prompt || original_description || description || name

Reading agent output:

task.agent_output_text     # reads TaskRun first, falls back to description regex
task.has_agent_output?     # checks TaskRun OR description pattern
task.latest_run            # most recent TaskRun
task.effective_prompt      # prompt that would be sent to agent
task.description_section("Agent Output")  # legacy fallback only

Key domains

• Task lifecycle: inbox -> up_next -> in_progress -> in_review -> done
• Factory loops: persistent automation cycles with logs/findings
• Swarm ideas: curated launch templates that create tasks
• ZeroBitch: external fleet/agent execution surface

6) Dev Commands

Setup + run

bin/setup
bin/dev
bin/rails server

DB

bin/rails db:prepare
bin/rails db:migrate
bin/rails db:reset

Tests + quality

bin/rails test
bin/rails test:system
bin/rubocop
bin/brakeman
bin/bundler-audit
bin/ci

7) Merge Gate Baseline (Do Not Skip)

Any change intended for integration should pass:

1. Lint/static checks (rubocop, security checks where applicable)
2. Unit/integration tests for touched areas
3. E2E/system tests for user-facing flows changed
4. Route/UI smoke check for impacted endpoints
5. Evidence written to docs/artifacts/ or docs/reports/

If any gate fails, do not mark task done.

8) Collaboration Rules for Agents

1. Edit code in the local clone only (G:\_OneDrive\OneDrive\Desktop\Py Apps\clawtrol\). Never SSH-edit /home/ggorbalan/clawdeck for non-emergency work — see .claude/rules/local-first-workflow.md.
2. Avoid broad refactors in dirty worktrees unless explicitly requested.
3. Never use destructive git commands (reset --hard, checkout --, force-push) unless asked.
4. When context is ambiguous, run the section 2 preflight before touching files.
5. Prefer small reversible changes with explicit validation steps.
6. Keep roadmap checkboxes updated in the same execution window.
7. Ship via /deploy-to-vm. Never pull source, build an image, or deploy a mutable tag on the VM.

9) Common Pitfalls and Fixes

Pitfall: "I changed files but app did not change"

Cause: either (a) edited the wrong folder (clawtrol-workspace/, clawdeck_remote_work/, or a gitclones/ reference clone), or (b) the exact tested SHA image was not deployed. Fix:

• Re-run section 2 preflight; confirm pwd ends in clawtrol and origin is wolverin0/clawtrol.
• Require green hosted CI, then run /deploy-to-vm <40-character-sha> and verify /health returns that SHA.

Pitfall: "Nav differs across pages"

Cause: duplicated nav partials drifting. Fix: update all nav partials together and verify desktop + mobile.

Pitfall: "Factory produced many commits but low trust"

Cause: missing merge gate/E2E evidence. Fix: enforce merge gate baseline before integration.

10) API/Auth Notes

• API base: /api/v1
• Auth header:
  • Authorization: Bearer <token>
• Agent identity headers (when used):
  • X-Agent-Name: <name>
  • X-Agent-Emoji: <emoji>

11) Related Docs

• docs/AGENT_INTEGRATION.md
• docs/OPENCLAW_INTEGRATION.md
• docs/API_REFERENCE.md
• docs/factory/ARCHITECTURE.md
• docs/roadmaps/

12) Swarm/Factory/ZeroClaw (Exception Mode Only)

Default execution mode is the normal task flow:

• One task at a time
• Standard board pipeline
• Normal review and validation gates

Do not activate Swarm/Factory/ZeroClaw by default.

Activate this mode only when at least one trigger is true:

1. The roadmap explicitly marks a phase as BURST, BATCH, or PARALLEL.
2. There are 5+ related tasks that can be safely parallelized.
3. A large refactor needs isolated playground validation before merge.
4. The user explicitly asks for swarm/factory/zeroclaw execution.

When exception mode is active:

• Swarm orchestrates phases and dispatch.
• Factory runs high-volume experiments in isolated workspaces.
• ZeroClaw handles heavy parallel subtasks and returns artifacts.

Guardrail:

• Never bypass the normal merge gates.
• If exception mode is not clearly required, stay on regular task flow.

Issue Tracking with bd (beads)

IMPORTANT: This project uses bd (beads) for ALL issue tracking. Do NOT use markdown TODOs, task lists, or other tracking methods.

Why bd?

• Dependency-aware: Track blockers and relationships between issues
• Git-friendly: Dolt-powered version control with native sync
• Agent-optimized: JSON output, ready work detection, discovered-from links
• Prevents duplicate tracking systems and confusion

Quick Start

Check for ready work:

bd ready --json

Create new issues:

bd create "Issue title" --description="Detailed context" -t bug|feature|task -p 0-4 --json
bd create "Issue title" --description="What this issue is about" -p 1 --deps discovered-from:bd-123 --json

Claim and update:

bd update <id> --claim --json
bd update bd-42 --priority 1 --json

Complete work:

bd close bd-42 --reason "Completed" --json

Issue Types

• bug - Something broken
• feature - New functionality
• task - Work item (tests, docs, refactoring)
• epic - Large feature with subtasks
• chore - Maintenance (dependencies, tooling)

Priorities

• 0 - Critical (security, data loss, broken builds)
• 1 - High (major features, important bugs)
• 2 - Medium (default, nice-to-have)
• 3 - Low (polish, optimization)
• 4 - Backlog (future ideas)

Workflow for AI Agents

1. Check ready work: bd ready shows unblocked issues
2. Claim your task atomically: bd update <id> --claim
3. Work on it: Implement, test, document
4. Discover new work? Create linked issue:
   • bd create "Found bug" --description="Details about what was found" -p 1 --deps discovered-from:<parent-id>
5. Complete: bd close <id> --reason "Done"

Auto-Sync

bd automatically syncs via Dolt:

• Each write auto-commits to Dolt history
• Use bd dolt push/bd dolt pull for remote sync
• No manual export/import needed!

Important Rules

• ✅ Use bd for ALL task tracking
• ✅ Always use --json flag for programmatic use
• ✅ Link discovered work with discovered-from dependencies
• ✅ Check bd ready before asking "what should I work on?"
• ❌ Do NOT create markdown TODO lists
• ❌ Do NOT use external issue trackers
• ❌ Do NOT duplicate tracking systems

For more details, see README.md and docs/QUICKSTART.md.

Landing the Plane (Session Completion)

When ending a work session, you MUST complete ALL steps below. Work is NOT complete until git push succeeds.

MANDATORY WORKFLOW:

1. File issues for remaining work - Create issues for anything that needs follow-up
2. Run quality gates (if code changed) - Tests, linters, builds
3. Update issue status - Close finished work, update in-progress items
4. PUSH TO REMOTE - This is MANDATORY:

   git pull --rebase
   bd sync
   git push
   git status  # MUST show "up to date with origin"

5. Clean up - Clear stashes, prune remote branches
6. Verify - All changes committed AND pushed
7. Hand off - Provide context for next session

CRITICAL RULES:

• Work is NOT complete until git push succeeds
• NEVER stop before pushing - that leaves work stranded locally
• NEVER say "ready to push when you are" - YOU must push
• If push fails, resolve and retry until it succeeds