tzachbon/srulik-toolkit / plugins/srulik-toolkit/skills/agent-swarm
agent-swarm
VerifiedCoordinate delegated work through the child-agent capabilities available in the current harness. Use when the user asks to spawn agents, delegate, swarm, parallelize, race approaches, or when active instructions require an orchestrator and worker split. Does not require Traycer, a specific provider, or a fixed model.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 14.0 KB |
| agents/openai.yaml | config | 208 B |
| graphs/pr-review.yaml | config | 1.1 KB |
| graphs/task.yaml | config | 8.8 KB |
| references/graph-spec.md | markdown | 19.5 KB |
| references/host-usage.md | markdown | 246 B |
| references/live-view.md | markdown | 2.6 KB |
| references/t3-code.md | markdown | 3.4 KB |
references/live-view.md
View on GitHub# Graph live view
Part of the [graph spec](graph-spec.md).
`graph.js view <run-dir> [--port N] [--host <address>]` serves a page that replays `run.jsonl` and streams new events. By default it listens on the machine's Tailscale IPv4 address (a `100.64/10` address on a `tailscale*` interface, or on a `utun*` interface that also has Tailscale's `fd7a:115c:a1e0::/48` address), so its link works from the owner's other devices, and on `127.0.0.1` when there is none. `--host` overrides it with one literal loopback or private address: `10/8`, `172.16/12`, `192.168/16`, Tailscale's `100.64/10`, or IPv6 `::1` or `fc00::/7`. Host names, public addresses and every spelling of `0.0.0.0` or `::` are refused, because the page has no login. It listens only on that address, not also on loopback. Requests whose `Host` header names any other address get 403, so a page on another site can't reach it through DNS rebinding.
The page draws the run as a node graph: one card per instance, with its item, a status word, the model and effort, the duration, its summary or error, and its attachments. Edges run from each instance's source; an instance that didn't fan out from one source gets an edge from every instance of each node it depends on. The graph is top-down by default, with a left-to-right toggle. The theme follows the system until the owner picks one with the theme button; both choices are kept in the browser. A header shows the graph name, its inputs, progress counts and elapsed time.
A parent card is marked `you` and shows `waiting on orchestrator` while open. A loop's later rounds get their own cards, labelled `r2`, `r3` and so on, with earlier rounds still visible. A skipped node shows a `skipped` card. An ordered item draws an edge from each item it waited for. The header shows each loop's round out of its maximum, and why it exited.
Clicking a card, or pressing Enter on it, opens a panel with the status, model, why it was picked, duration, inputs, error, output file path, attachments and output. Output, attachments and the run index render readably: JSON as labelled rows and cards, Markdown (also inside JSON strings) as headings, lists, tables and code, built as DOM text nodes, never parsed as HTML; links open only for http(s). A collapsed Raw section keeps the original text. Escape closes it. The Run index button shows `nodes/index.md`. The page reads those files from `/nodes/index.md`, `/nodes/<node>-<index>[.r<n>].md` and `/nodes/<node>-<index>[.r<n>]/<name>`. These are served as plain text with `nosniff`, and only when the real path stays inside `<run-dir>/nodes`; any other path gets 404.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |