MCP App StoreMCP App Store
from modal-auto-research-skills21
Orchestrate multiple autonomous Claude Code agents across separate GPUs or sandboxes to run parallel experiments, debugging sessions, or batch workloads with st
This skill provides an orchestration pattern to spawn and manage multiple autonomous Claude Code agents, each running in its own GPU-backed sandbox or as a lightweight experiment caller. It describes deployment best-practices (deploy once, call many), interactive SSH-driven sandboxes for debugging, and non-interactive batch experiment patterns. The skill includes tools for agents to report structured findings and for the parent process to monitor progress and read trajectories or summaries.
Use this skill when you need to run parallel experiments (hyperparameter sweeps, model comparisons), concurrent debugging sessions across GPUs, or distributed data-processing agents that must report structured results back to a coordinator. It is ideal for workloads where isolation per-agent (separate sandbox/VM/GPU) improves reliability or throughput.
references/sub-agents.md with full orchestration guidance.Primarily designed for Claude Code / Claude-based subprocess orchestration; patterns apply to other LLM agent runtimes that support external process invocation and remote sandboxes (e.g., Sonnet, custom Claude CLI integrations).
Sub-agents is a parallel Claude Code orchestration skill for spawning multiple autonomous agents across separate GPUs via Modal. SKILL.md is well-structured with clear quick reference and two deployment options (SSH sandbox and batch). No bundled scripts to test. Notable concerns: instructs agents to use --dangerously-skip-permissions and disables SSH host verification, both significant security trade-offs. Requires Modal platform and modal-gpu-dev skill as dependencies.
The --dangerously-skip-permissions flag is concerning from a security standpoint as it instructs spawned agents to bypass all permission checks. The SSH host key bypass is a convenience trade-off but reduces security. Skill is functional and well-documented but depends heavily on the Modal ecosystem.
Yielding Bear — Unified LLM Routing
A unified LLM routing API that automatically routes requests to the cheapest capable model across providers to reduce inference cost by ~60-80%.
AWP (Agent Work Protocol)
Tooling and scripts for onboarding, staking, allocation, and managing agents on the AWP network (Base/Ethereum/Arbitrum/BSC). Includes safe, opt-in daemon and r
Starlark Dev
Create, debug, and test Kurtosis Starlark packages — write package structure, run dry-runs, and inspect plan execution for reliable orchestration.
Maestro Orchestrate
Orchestrates complex engineering workflows with enforced design and approval gates, child-agent delegation, and session state management.
Run Execute
Orchestrates execution of work items across modes (autopilot, confirm, validate) with scripted init/complete tooling, plan/test/report artifacts and strict gati
Cost Tracker
Monitor agent session costs, set budget alerts, and get actionable token-spend optimizations to keep multi-session workflows within budget.
Canary — Post-Deploy Visual Monitor
Run a short post-deploy monitor that captures screenshots, checks console errors, and compares performance against baselines to detect regressions and page fail
Overnight — Autonomous Long-Running Coding
Orchestrates long-running coding goals: decomposes objectives into atomic tasks, dispatches isolated worktree workers, verifies acceptance criteria, and merges