Comfy-Cozy: Command for Claude Code

.claude/commands/PRODUCTION_HARDEN.md

PRODUCTION_HARDEN is a command for Claude Code from JosephOIbrahim/Comfy-Cozy. It costs 0 tokens per session (3,513 once invoked), scanned A, original, MIT.

A production-hardening workflow that coordinates several specialized coding agents to inspect and improve a ComfyUI-based VFX application. Each agent focuses on an area such as visual workflow correctness, API compatibility, pipelines, operations, or systems engineering.

In plain words
What is it for?
Use it to harden the target VFX workstation application, run specialized reviews, and organize proposed fixes as conventional-commit pull requests from separate agent branches.
Why use it?
It divides a large reliability review into domain-specific investigations and combines the results through branches and pull requests. This helps teams examine performance, security, testing, packaging, and integration concerns together.

Command for Claude Code

Written for Claude Code: installed under .claude/. Also seen: reads .claude/ paths; mentions CLAUDE.md; mentions subagents.

This is JosephOIbrahim/Comfy-Cozy's own configuration. It tells Claude Code how to work on Comfy-Cozy itself, so it is not a mod to install elsewhere. Copy it as a starting point and replace the rules that are about this project. Everything Comfy-Cozy configures →

Needs its repository: it runs a file that does not travel with it, so clone the repository first. The line is python .claude/agents/orchestrate.py --workstream WS-1 --agent SYS_ENG.

Reuse

Borrowing it

Nothing to install: this file belongs to JosephOIbrahim/Comfy-Cozy. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.

Copy the file
curl -O https://raw.githubusercontent.com/JosephOIbrahim/Comfy-Cozy/master/.claude/commands/PRODUCTION_HARDEN.md
Clone the repo
git clone --depth 1 https://github.com/JosephOIbrahim/Comfy-Cozy

Made for: Claude Code.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for PRODUCTION_HARDEN

README.md
[![agentmods](https://agentmods.dev/badge/commands/josephoibrahim/comfy-cozy/production_harden/github.svg)](https://agentmods.dev/commands/josephoibrahim/comfy-cozy/production_harden)
Your own site
<a href="https://agentmods.dev/commands/josephoibrahim/comfy-cozy/production_harden"><img src="https://agentmods.dev/badge/commands/josephoibrahim/comfy-cozy/production_harden/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for PRODUCTION_HARDEN

Your own site · 80×15
<a href="https://agentmods.dev/commands/josephoibrahim/comfy-cozy/production_harden"><img src="https://agentmods.dev/badge/commands/josephoibrahim/comfy-cozy/production_harden.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 0 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 3,513 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00000 $0.03513
Opus 5 $0.00000 $0.01757
Sonnet 5 $0.00000 $0.00703
Haiku 4.5 $0.00000 $0.00351

Measured 11d ago against content hash aa98b4fe18ba, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-11, from the pricing page.

Security

Grade A, and why

PRODUCTION_HARDEN scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 11d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

.claude/commands/PRODUCTION_HARDEN.md · 280 lines

How it starts

The opening of the file, as written. The whole thing — 280 lines — stays where its author put it; the contents beside it link to each section on GitHub.

PRODUCTION HARDENING — MoE Agent Team Orchestration

Mission: Autonomously production-harden ComfyUI-Comfy Cozy Agent using specialized agent teams routed by Mixture-of-Experts domain decomposition.

Runtime: Claude Code with Sonnet 4.6 sub-agents Target: VFX workstation — Threadripper PRO 7965WX, RTX 4090, 128GB DDR5, Windows Baseline: 497 tests passing, 0 lint errors, 10,678 LOC agent / 6,863 LOC tests


MoE TEAM ROSTER

Each agent is a Claude Code sub-agent invoked via claude --model claude-sonnet-4-6-20250929 with a role-specific prompt file. Agents operate on branches, submit PRs via conventional commits.

Agent Role Domain Branch Prefix
VFX_SUPER VFX Supervisor Artistic intent, UX, workflow correctness, naming harden/vfx-
COMFY_LEAD ComfyUI Team Lead ComfyUI API contracts, node compatibility, MCP protocol harden/comfy-
NUKE_COMP Nuke Compositor Pipeline integration, I/O patterns, cross-app interop harden/pipeline-
PRODUCER Producer CI/CD, packaging, release, docs, metrics, scheduling harden/ops-
SYS_ENG Systems Engineer Performance, concurrency, security, error handling, typing harden/sys-

Routing Rules (MoE Gating)

Task mentions UI/UX/artist experience    → VFX_SUPER
Task mentions ComfyUI API/nodes/MCP      → COMFY_LEAD
Task mentions I/O/pipeline/integration   → NUKE_COMP
Task mentions CI/docs/packaging/release  → PRODUCER
Task mentions perf/security/types/async  → SYS_ENG
Task spans multiple domains              → Primary + Review by secondary

AND-NODE TASK TREE

All branches must succeed. Hardest branches marked with ⚠️.

PRODUCTION HARDEN (AND — all required)
├── WS-1: Type Safety & Static Analysis (SYS_ENG) ⚠️ HARDEST
│   ├── 1.1 Add py.typed marker + pyright config
│   ├── 1.2 Type-annotate all public APIs (agent/tools/*.py)
│   ├── 1.3 Type-annotate brain layer (agent/brain/*.py)
│   ├── 1.4 Type-annotate core (config, circuit_breaker, rate_limiter, etc.)
│   ├── 1.5 Add pyright to CI (strict mode on new code, basic on existing)
│   └── 1.6 Fix all pyright errors at basic level
│
├── WS-2: Async Architecture Hardening (SYS_ENG + COMFY_LEAD review)
│   ├── 2.1 Audit all sync-in-async bridges (run_in_executor patterns)
│   ├── 2.2 Add proper async context managers for httpx clients
│   ├── 2.3 Fix MCP server graceful shutdown (signal handling)
│   ├── 2.4 Add connection pooling for ComfyUI HTTP calls
│   ├── 2.5 Add timeout configuration (per-tool configurable)
│   └── 2.6 Add async health check endpoint for MCP server
│
├── WS-3: Error Handling & Resilience (SYS_ENG + VFX_SUPER review)
│   ├── 3.1 Audit all exception handlers — no silent swallows
│   ├── 3.2 Add structured error types (ToolError, TransportError, ValidationError)
│   ├── 3.3 Add error context propagation (chain exceptions properly)
│   ├── 3.4 Add circuit breaker metrics (open/close counts, latency p50/p99)
│   ├── 3.5 Add retry budget (prevent infinite retry storms)
│   └── 3.6 Add graceful degradation paths (ComfyUI down → partial functionality)
│
├── WS-4: State Management Hardening (COMFY_LEAD + SYS_ENG review) ⚠️
│   ├── 4.1 Audit all module-level mutable state (6 _instance singletons, _state, _cache)
│   ├── 4.2 Add session-scoped state isolation (remove workflow_patch._state global)
│   ├── 4.3 Add session validation on load (schema versioning, migration)
│   ├── 4.4 Add atomic file writes for sessions (write-tmp + rename)
│   ├── 4.5 Add state cleanup / garbage collection for stale sessions
│   └── 4.6 Add memory limits for in-process caches (_cache in comfy_discover)
│
├── WS-5: Security Hardening (SYS_ENG + NUKE_COMP review)
│   ├── 5.1 Audit path traversal protection (validate_path coverage)
│   ├── 5.2 Add input validation for all MCP tool inputs (jsonschema strict)
│   ├── 5.3 Add rate limiting for MCP tool calls (prevent abuse)
│   ├── 5.4 Add secrets handling audit (.env, API keys, no hardcoded values)
│   ├── 5.5 Add dependency vulnerability scan (pip-audit in CI)
│   └── 5.6 Docker security review (non-root, no unnecessary capabilities)
│
├── WS-6: Testing Hardening (COMFY_LEAD + SYS_ENG)
│   ├── 6.1 Fix Pillow deprecation warnings (getdata → get_flattened_data)
│   ├── 6.2 Add property-based tests for patch engine (hypothesis)
│   ├── 6.3 Add integration test harness (requires live ComfyUI — mark as slow)
│   ├── 6.4 Add test coverage measurement + enforce minimum (pytest-cov)
│   ├── 6.5 Add mutation testing on critical paths (mutmut on workflow_patch)
│   ├── 6.6 Add fuzz testing for workflow JSON parsing
│   └── 6.7 Add MCP protocol conformance tests
│
├── WS-7: Observability & Monitoring (PRODUCER + SYS_ENG)
│   ├── 7.1 Add structured metrics (tool call counts, latencies, error rates)
│   ├── 7.2 Add tool call tracing (correlation ID → tool chain visualization)
│   ├── 7.3 Add performance baselines (benchmark suite for key operations)
│   ├── 7.4 Add memory profiling for long-running MCP sessions
│   ├── 7.5 Add log aggregation-friendly format (OpenTelemetry-compatible)
│   └── 7.6 Add GPU utilization monitoring hooks (RTX 4090 specific)
│
├── WS-8: CI/CD & Packaging (PRODUCER)
│   ├── 8.1 Add Python 3.13 to CI matrix
│   ├── 8.2 Add pyright/mypy check to CI
│   ├── 8.3 Add pip-audit (dependency security) to CI
│   ├── 8.4 Add test coverage gate to CI (fail below threshold)
│   ├── 8.5 Add docker-compose.yml for local dev (ComfyUI + agent)
│   ├── 8.6 Add release automation (version bump, changelog, PyPI publish)
│   ├── 8.7 Add pre-commit hooks config (.pre-commit-config.yaml)
│   └── 8.8 Add Windows-specific CI validation (paths, comfy_cli integration)
│
├── WS-9: Documentation & UX (VFX_SUPER + PRODUCER)
│   ├── 9.1 Add CONTRIBUTING.md (developer onboarding)
│   ├── 9.2 Add CHANGELOG.md (keepachangelog format)
│   ├── 9.3 Update README for production deployment
│   ├── 9.4 Add troubleshooting guide (common errors, ComfyUI connection issues)
│   ├── 9.5 Add MCP tool reference (auto-generated from TOOLS schemas)
│   ├── 9.6 Add architecture diagram (mermaid in docs/)
│   └── 9.7 Audit all user-facing error messages for artist-friendly language
│
├── WS-10: Platform & Config Hardening (NUKE_COMP + SYS_ENG)
│   ├── 10.1 Add config validation on startup (fail fast with clear errors)
│   ├── 10.2 Add platform-aware defaults (Windows/macOS/Linux path handling)
│   ├── 10.3 Add ComfyUI auto-discovery (find running instance, check common ports)
│   ├── 10.4 Add comfy_cli integration (leverage local ComfyUI management)
│   ├── 10.5 Add environment variable documentation (.env.example)
│   └── 10.6 Hardware profile auto-detection (GPU model, VRAM, for optimizer hints)
│
└── WS-11: MCP Protocol Hardening (COMFY_LEAD) ⚠️
    ├── 11.1 Add MCP resource support (expose workflow state as MCP resources)
    ├── 11.2 Add MCP prompt support (pre-built prompt templates)
    ├── 11.3 Add SSE transport option (for remote/web clients)
    ├── 11.4 Add MCP server versioning (protocol negotiation)
    ├── 11.5 Add tool schema validation (ensure all schemas are MCP-compliant)
    └── 11.6 Add MCP error code compliance (proper JSON-RPC error responses)

Read the full file on GitHub · 280 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 11d ago First seen · 280 lines · 0 tokens per session scan A aa98b4fe18ba

Subscribe to this mod's changes

PRODUCTION_HARDEN is a command published in the GitHub repository JosephOIbrahim/Comfy-Cozy (24 stars, last pushed 1mo ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 3,513 tokens. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.