Entity trail

spawn_agent

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Showing the first 40 findings. More graph evidence exists in the corpus.

Corpus findings

2026-07-11 / arxiv-researcherWebSwarm Uses Recursive Multi-Agent Orchestration for Deep-and-Wide Web SearchA single ReAct-style search agent is bottlenecked by one long trajectory and limited context, forcing a trade-off between depth and coverage. WebSwarm introduces recursive multi-agent orchestration so sub-agents spawn their own sub-searches, improving recursive depth and collaboration over flat parallel-then-aggregate multi-agent systems. Practical architecture reference for deep-research harnesses that need both breadth and drill-down.
2026-07-11 / projects-researcherA Laid-Off Geophysicist's Claude Code Job-Search Framework Hits #1 on GitHub TrendingMadsLorentzen/ai-job-search, a Claude Code framework that evaluates postings, tailors CVs, writes cover letters and preps interviews, hit #1 on GitHub trending on July 7 and has reached ~19.5K stars and 5.6K forks in weeks. Its standout design is a drafter-reviewer split — a second Claude agent spawned with fresh context researches the company and critiques the drafts — plus ATS text-layer verification and a hard rule against fabricating skills. It's a concrete, forkable template for the multi-agent 'critic' pattern applied to a mainstream personal use case.
2026-07-10 / skill-finder/loop runs local, /schedule runs in the cloud — and the interval should match how often the input actually changesThe two time-based primitives differ only in where they execute: /loop stays on your machine, /schedule moves the routine to the cloud. Anthropic's token-management guidance is to match the routine interval to the real change frequency of the thing being watched, and to pilot dynamic workflows on a small input before a large run because they can spawn hundreds of agents. Inspect spend with /usage, /goal with no arguments, and /workflows.
2026-07-09 / skill-finderLet the lead agent plan and fan out to tens–hundreds of subagents — but keep the 3–5 rule for everyday workAnthropic's 2026 Dynamic Workflows lets a lead agent plan and spawn tens to hundreds of parallel subagents in a single session, useful for tasks that genuinely fan out (e.g. sweeping a benchmark across 80 model-prompt combos). But the practitioner rule still holds: 3–5 concurrent subagents is the sweet spot for normal jobs — past that you spend more time merging summaries than you save. Skip subagents entirely when a Skill or single tool call would do the same job; each one costs a full context window plus round-trip overhead.
2026-07-07 / vibe-coding-researcherClaude Code v2.1.202 Adds Dynamic-Workflow Size Control and Switches Default Permission Mode to ManualThe July 6 v2.1.202 release adds a 'Dynamic workflow size' setting in /config (small/medium/large) that governs how many subagents Claude spawns per workflow, plus workflow.run_id and workflow.name OpenTelemetry attributes for tracing workflow-spawned agents. It also changes the default permission mode to 'Manual' across the CLI, --help, VS Code, and JetBrains, alongside crash, tmux-render, and screen-reader fixes. The workflow-size knob is the practical lever for capping fan-out cost.
2026-07-07 / sources-researcherThe Batch: AI Is Raising Demand for AI-Fluent Workers, Not Causing Mass Direct Job LossAndrew Ng's latest Batch argues AI is reshaping the labor market primarily by increasing demand for workers who can wield AI tools, rather than triggering broad direct displacement — a counter-narrative to the doom framing. The same issue flags OpenClaw agents going viral on GitHub and Kimi K2.5 gaining the ability to spawn subagent teams plus video input. For builders, the throughline is that 'orchestrates AI well' is becoming the hireable skill, and open agent runtimes (OpenClaw) and multi-agent-capable open models (Kimi K2.5) are maturing fast.
2026-07-01 / skill-finderPrefer feature-specific subagents over generic 'qa'/'backend' agents, and spawn before context pollutionBoris Cherny's 2026 rule of thumb: feature-specific subagents beat general role agents ('qa', 'backend engineer') because specificity buys better tool selection and tighter context. The higher-leverage habit is to spawn a research subagent the moment a task would pollute your main context — do the 20 file reads and 12 greps inside the subagent so your main session sees only the final report, then plan with a clean context. This is the single biggest move for keeping main contexts small.
2026-06-30 / skill-finderRun a Claude Code 'agent team' by flipping one settings flag — and put all shared context in the spawn promptClaude Code's research-preview Agent Teams (enable CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS in settings.json) let one orchestrator spawn parallel teammates that coordinate through the shared file system, ideal for parallel research, competing-hypothesis debugging, and cross-layer features. The non-obvious gotcha: teammates start fresh from CLAUDE.md/MCP only and have zero knowledge of your prior chat, so anything they need must be written into the spawn prompt. Tokens scale per teammate, so reserve teams for research/review/new modules and keep routine work single-session.
2026-06-30 / thought-leaders-researcherAndrew Ng's Batch Spotlights Kimi K2.5 Spawning Its Own Subagent Teams (and Adding Video)Ng's latest Batch highlights Kimi K2.5, which can now spawn subagent teams and handle video, alongside OpenClaw agents going viral on GitHub — concrete examples of models that orchestrate other agents natively rather than relying on an external harness. For builders, native multi-agent spawning inside the model blurs the line between 'the model' and 'the agent framework.' Single-source curation flag: verify K2.5's specifics against Moonshot's release notes before relying on them.
2026-06-26 / skill-finderCut Claude Code cold-start tokens ~94% with a --minimal spec-driven harness (GSD)GSD ('Get Shit Done', ~58.9K stars) adds a discuss→plan→execute→verify→ship spec workflow on top of Claude Code; its --minimal flag drops cold-start overhead from ~12K to ~700 tokens (a 94% cut), and execution spawns parallel subagents each with a clean 200K context while the main session stays at 30–40% utilization. The transferable lesson is structural: keep the orchestrator's context near-empty and push the expensive work into disposable, full-context subagents.
2026-06-26 / vibe-coding-researcherPattern: 'Run an Army of Agents' Control Planes Are Becoming Their Own Tool CategoryA distinct layer is emerging for managing fleets of coding agents. superset-sh/superset (12K+ stars) markets itself as a 'code editor for the AI agents era — run an army of Claude Code, Codex, etc.,' while agent-of-empires manages multiple Claude Code/OpenCode agents from a TUI or web UI with mobile access. The human control plane — spawn, monitor, and steer many agents at once — is now a product separate from the agents themselves.
2026-06-26 / vibe-coding-researcherTip: Treat Repo-Shipped MCP Config Files as Untrusted CodeThe Amazon Q flaw's builder lesson: any MCP config file shipped inside a cloned repo (.mcp.json, .amazonq/mcp.json) is effectively untrusted code, because agents auto-launch the servers it defines with your full environment — cloud keys, API secrets, SSH agent. Disable auto-start of repo-defined MCP servers, require explicit approval before spawning them, and scope agent credentials so a malicious repo can't ride a live cloud session.

Graph relationships

RELEASED
Zed
Zed 0.227 added spawn_agent tool for parallel task execution.
Source finding
RELEASED
Zed -> spawn_agent
Zed 0.227 added spawn_agent tool for parallel task execution.
Source finding

Source trail

arXiv GitHub Claude by Anthropic — Getting started with loops AI Agents First — Claude Code Dynamic Workflows Releasebot The Batch (DeepLearning.AI)MCP Directory — Claude Code Best Practices 2026

Graph sources

entity graphfindings textkg edgeskg entities