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.
npx skills add ulises-jeremias/agent-toolkit --skill herdrgit clone --depth 1 https://github.com/ulises-jeremias/agent-toolkitWrote 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.
[](https://agentmods.dev/skills/ulises-jeremias/agent-toolkit/herdr)<a href="https://agentmods.dev/skills/ulises-jeremias/agent-toolkit/herdr"><img src="https://agentmods.dev/badge/skills/ulises-jeremias/agent-toolkit/herdr/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.
<a href="https://agentmods.dev/skills/ulises-jeremias/agent-toolkit/herdr"><img src="https://agentmods.dev/badge/skills/ulises-jeremias/agent-toolkit/herdr.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00029 | $0.01389 |
| Opus 5.5 | $0.00012 | $0.00556 |
| Sonnet 5.5 | $0.00006 | $0.00278 |
| Haiku 4.5 | $0.00003 | $0.00139 |
Grade A, and why
herdr 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 1mo 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.
How it starts
The opening of the file, as written. The whole thing — 110 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Herdr
Operate Herdr as the primary UI for agent-toolkit swarm — eager tabs, shell-consistent execution (zsh/bash via _user_shell()), and clean reuse of windows. This skill is the Herdr complement to swarm and swarm-observer.
When to use
- Swarm was started with
--ui herdr(default) and you need to inspect or recover Herdr state. herdr workspace list/create/tab create/pane runbehavior is unexpected (extra workspacesw7-wAforheadless, wrong--cwd, or missing eager windows).- User prefers Herdr over tmux for its excellent interactive UX and wants to reuse tabs for sequential requests.
Prerequisites
herdrCLI installed and daemon running (herdr --version,herdr workspace list --jsonworks;--jsonmay be unsupported on older binaries — fallback to text).- Swarm run exists:
.agent-toolkit/swarm/runs/<run-id>/state.jsonwithui: herdr. - Repo is a Git directory (worktrees are created under
.agent-toolkit/swarm/runs/<run-id>/worktrees/).
Workflow
1. Verify backend and shell
agent-toolkit swarm doctor --json | jq '.backends'
herdr --version
echo $SHELL; getent passwd $USER | cut -d: -f7 # should be /usr/bin/zsh on this host
# swarm's _user_shell() picks: $SHELL -> pwd.getpwuid -> /usr/bin/zsh -> /bin/bash
2. Inspect Herdr state for a run
herdr workspace list --json 2>&1 | jq '.[] | {id,name,path,status}' | head -n 40
# Expected: one workspace per run: swarm-20260807T...-<short>
# Orphan check: workspaces w7-wA with headless run-id -> delete after fixing backend
agent-toolkit swarm status --run-id <run-id> --json | jq '{ui,recipe,runner,model_profile,worktrees,handoffs}'
agent-toolkit swarm handoffs --run-id <run-id> --json | jq
herdr workspace list | grep swarm-<run-id>
3. Eager windows (automatic on swarm start)
On swarm start --ui herdr, swarm creates one tab per role immediately:
tab create --workspace swarm-<run-id> --name <role>for each role (implementer,reviewer,integrator, ... plus1for coordinator).- Each non-implementer tab runs:
echo "Waiting for handoff: <pred> -> <role> | role: <role> | run: <run-id>"via<shell> -lc(quoted to avoid| pipeissues) and recordsagent_waitingtrace. select-window/workspace focusfocusesimplementer; implementer pane runsexport AGENT_TOOLKIT_SWARM_RUN_ID=... RUN_DIR=... REPO=... SWARMFORGE_ROLE=implementer && cd <worktree> && exec <runner> "$(cat prompt)".
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.
- 1mo ago First seen · 110 lines · 29 tokens per session scan A 3bad25c18389
herdr is a skill published in the GitHub repository ulises-jeremias/agent-toolkit (19 stars, last pushed yesterday), licensed MIT. It adds 29 tokens to every session and 1,389 once invoked, about $0.0001 per session on Opus 5.5. 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.
Other skills, from other repositories
collection
Choose which SkillNote skill collection is active for this Codex project. Use when the user says "change collection", "switch skills", "use frontend skills", or "show collections".
audit-accomplishments
Use when preparing a performance self-reflection, midyear or annual review, promotion packet, or brag document; when asked to collect, mine, gather, or summarize your accomplishments, contributions, or impact over a period; or when assembling cited evidence of work done across GitHub, Jira, Notion, Slack, and local…
start-day
Use at the start of a workday or when the user says "start my day", "good morning", "/start-day", "morning", "what should I work on today", or wants a catch-up on what happened while they were away. Use when the user wants a short ranked list of what to tackle today.
tracking-time
Tracks time on tasks using timers or manual entries and generates timesheet reports in Astravue. Use when logging work hours, starting or stopping a timer, reviewing time entries, generating timesheets, or asking "how much time was spent on this?".
local-mcp-daily-brief
Use when the user asks "what's on my plate today", "catch me up", "morning brief", or wants a summary of their day across calendar, reminders and email. The single most common multi-step automation. Powered by LMCP on macOS and Windows.
local-mcp-email-to-action
Use when the user wants to deal with their inbox — find what needs a response, draft replies, and turn emails into calendar events or reminders. Captures the common email→read→reply→schedule workflow on macOS (Apple Mail, Calendar, Reminders). Powered by LMCP.