swarm-observer

swarm-observer is a skill for Claude Code, Codex from ulises-jeremias/agent-toolkit. It costs 32 tokens per session (1,334 once invoked), scanned A, original, MIT.

A monitoring and recovery guide for Agent Toolkit swarm runs, where several coding agents work in parallel. It checks their status, handoffs, logs, worktrees, and shells to find stalled or failed work.

In plain words
What is it for?
Use it to inspect swarm progress, view logs, attach to a run, or recover stuck workspaces and handoffs.
Why use it?
It makes it easier to tell whether a swarm is still working, stuck, or finished. It helps diagnose handoff failures and recover a run without losing its agent windows.

Skill for Claude CodeCodex

Written for no agent in particular: nothing here depends on one. Also seen: mentions OpenCode.

Part of the agent-toolkit-complete plugin — 116 skills shipped together

Good fit Use it to inspect swarm progress, view logs, attach to a run, or recover stuck workspaces and handoffs.

Compare 6 skills from other repositories ↓
Install with agentmods
npx agentmods add skills/ulises-jeremias/agent-toolkit/swarm-observer
Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

Any agent
npx skills add ulises-jeremias/agent-toolkit --skill swarm-observer
Clone the repo
git clone --depth 1 https://github.com/ulises-jeremias/agent-toolkit

Made for: Claude Code, Codex.

Or install agent-toolkit-complete, the plugin that ships this one along with the rest of its 116 skills.

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 swarm-observer

README.md
[![agentmods](https://agentmods.dev/badge/skills/ulises-jeremias/agent-toolkit/swarm-observer/github.svg)](https://agentmods.dev/skills/ulises-jeremias/agent-toolkit/swarm-observer)
Your own site
<a href="https://agentmods.dev/skills/ulises-jeremias/agent-toolkit/swarm-observer"><img src="https://agentmods.dev/badge/skills/ulises-jeremias/agent-toolkit/swarm-observer/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 swarm-observer

Your own site · 80×15
<a href="https://agentmods.dev/skills/ulises-jeremias/agent-toolkit/swarm-observer"><img src="https://agentmods.dev/badge/skills/ulises-jeremias/agent-toolkit/swarm-observer.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 32 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,334 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. Third-party audits
  • NVIDIA SkillSpector pass 7 Sept 2026
How audits are shown
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 $0.00032 $0.01334
Opus 5 $0.00016 $0.00667
Sonnet 5 $0.00006 $0.00267
Haiku 4.5 $0.00003 $0.00133

Measured 5d ago against content hash e1f29e946fdb, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-08, from the pricing page.

Security

Grade A, and why

swarm-observer 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 5d 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.

plugins/agent-toolkit-complete/.github/skills/swarm-observer/SKILL.md · 103 lines

How it starts

The opening of the file, as written. The whole thing — 103 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Swarm Observer

Monitor any agent-toolkit swarm run, diagnose stuck handoffs or backend drift, and recover without losing windows. Works for both Herdr (tabs) and tmux (isolated socket agent-toolkit-swarm-<run-id>). Integrates with swarm's eager-window model (every role starts as Waiting for handoff: <pred> -> <role> with agent_waiting trace).

When to use

  • A swarm was launched and the user asks "is it done?", "what's the status?", "show logs", or "attach".
  • A handoff has been active for >5m (955be6f6e2c... pattern) or worktree_failed.
  • Herdr shows extra workspaces (w7-wA) for headless runs — the get_backend("headless") fallback case.
  • User wants to reuse the same run for a new request and needs to know if windows are still alive.

Prerequisites

  • Run lives under .agent-toolkit/swarm/runs/<run-id>/state.json (filesystem is authoritative).
  • Know run-id (20260807T...-xxxxxx) from swarm start --json or swarm list --json.
  • Shell is detected (_user_shell(): $SHELLpwd.getpwuid/usr/bin/zsh fallback, executed as <shell> -lc).

Workflow

1. List and pick the run

agent-toolkit swarm list --json | jq
agent-toolkit swarm status --run-id <run-id>
agent-toolkit swarm status --run-id <run-id> --json | jq '.handoffs,.worktrees,.trace'

2. Inspect handoffs and trace

agent-toolkit swarm handoffs --run-id <run-id>
agent-toolkit swarm handoffs --run-id <run-id> --json | jq '.[] | {id,from,to,status,artifact,auto}'
# Active that should have auto-completed:
#   where to==from_role and handoff create --to == that role -> auto:true

Trace keys to watch: agent_waiting, handoff_completed auto:true, worktree_failed, headless_fallback.

3. Tail logs and attach

agent-toolkit swarm logs --run-id <run-id> --follow
agent-toolkit swarm logs --run-id <run-id> --role reviewer --follow
agent-toolkit swarm attach --run-id <run-id>          # herdr: workspace focus; tmux: attach -L agent-toolkit-swarm-<run-id>
herdr workspace list --json 2>&1 | jq '.[] | {id,name,status}'
tmux -L agent-toolkit-swarm-<run-id> capture-pane -t <window> -p | tail -n 80

Read the full file on GitHub · 103 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. 5d ago First seen · 103 lines · 32 tokens per session scan A e1f29e946fdb

Subscribe to this mod's changes

swarm-observer is a skill published in the GitHub repository ulises-jeremias/agent-toolkit (16 stars, last pushed today), licensed MIT. It adds 32 tokens to every session and 1,334 once invoked, about $0.0002 per session on Opus 5. 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 2026-09-04.

Related

Other skills, from other repositories

skillnote-doctor

A diagnostic tool for OpenClaw agents -- checks skill registry connectivity, AGENTS.md setup, config file validity, and installed skill health. Use when your setup seems broken, skills aren't loading, or you want to audit your agent's configuration.

luna-prompts/skillnote · 54 tokens

chatcrystal-debug-recall

Recall ChatCrystal memories for debugging tasks involving failing tests, compiler errors, runtime exceptions, dependency issues, environment breakage, or performance regressions. Use when historical root causes, fixes, or pitfalls may accelerate diagnosis before proposing a fix.

ZengLiangYi/ChatCrystal · 54 tokens

hotfix

Fixes an observed defect with reproducible evidence in one call: writes a short trace doc before touching code, implements the fix, and backs it with a regression test written before the fix. Production incidents are the motivating case, not a gate. When blocked, it halts by name and saves the doc for a later call to…

pe-menezes/vibeflow · 102 tokens

iblai-api-agent-audit

Read an ibl.ai agent's audit log via the platform API — a who-changed-what trail filterable by action, actor email, and date range. Read-only. Use when investigating configuration changes to an agent.

iblai/api · 52 tokens

analyze-failures

Triage Katalon True Platform/TestOps test failures and file defects. Use when you need to investigate failed test results, classify each failure as product defect vs automation defect vs environment/data issue, cluster failures by common signature, find likely root cause from execution data, and optionally create…

katalon-labs/true-skills · 131 tokens

bisect

Use when hunting a regression, phrases like "bisect", "find the commit that broke X", "this used to work", "regression in test Y", "when did start". Also use when escalated from ci-debug-loop because log analysis can't pinpoint the offending change, or when a previously-passing test/build/behavior is now failing and…

paultyng/skill-issue · 121 tokens