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 skrrt-sh/skills --skill herdr-agentsgit clone --depth 1 https://github.com/skrrt-sh/skillsWrote 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/skrrt-sh/skills/herdr-agents)<a href="https://agentmods.dev/skills/skrrt-sh/skills/herdr-agents"><img src="https://agentmods.dev/badge/skills/skrrt-sh/skills/herdr-agents/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/skrrt-sh/skills/herdr-agents"><img src="https://agentmods.dev/badge/skills/skrrt-sh/skills/herdr-agents.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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 | $0.00116 | $0.01311 |
| Opus 5.5 | $0.00046 | $0.00524 |
| Sonnet 5.5 | $0.00023 | $0.00262 |
| Haiku 4.5 | $0.00012 | $0.00131 |
Grade A, and why
herdr-agents scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 100 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Herdr Agents
The subagents skill decides the scopes. This skill runs each scope in its own checkout.
Guard: test "${HERDR_ENV:-}" = 1. If it fails, say so and dispatch with the Agent tool.
Read-only Explore work always uses the Agent tool.
The bundled herdr-mail plugin holds mail until the recipient is idle. Its hooks mail each
child's done, blocked, and exited states to the parent. Check it once per run:
mail="${CLAUDE_SKILL_DIR}/assets/herdr-mail/bin/herdr-mail"
herdr plugin list --json | jq -e --arg r "${CLAUDE_SKILL_DIR}/assets/herdr-mail" \
'.result.plugins[] | select(.plugin_id == "skrrt.herdr-mail" and .plugin_root == $r)'
No match: ask the user, then run herdr plugin link "${CLAUDE_SKILL_DIR}/assets/herdr-mail". It
runs in every Herdr session. Suggest allowing Bash(<$mail path>:*), or each child stops on a
permission prompt at its first send.
Dispatch
Once per run, confirm herdr worktree create, herdr agent start, and $mail are allowed. If
not, ask the user once for the rules, not per agent.
Once per stage, snapshot the base: base=$(git stash create), or HEAD on a clean tree. It
captures the integrated earlier stages without committing or signing.
Per scope. Name the agent after the scope, matching [a-z][a-z0-9_-]{0,31}. Read every ID from
JSON.
ws=$(herdr worktree create --cwd "$PWD" --branch "scope/$name" --base "${base:-HEAD}" --label "$name" --no-focus)
pane=$(jq -r .result.root_pane.pane_id <<<"$ws") # also keep .result.workspace.workspace_id
herdr agent start "$name" --kind claude --pane "$pane" -- --model opus
"$mail" watch "$name"
herdr agent prompt "$name" "$(cat <<'EOF'
<brief, ending with the protocol below>
EOF
)" --wait --until working --timeout 30000
--until workingconfirms delivery and returns. Start the whole stage before waiting on any agent.- Use another kind, model, or permission mode only when the user names one.
- Install dependencies in the checkout before prompting. Get its path from
herdr worktree list --cwd "$PWD". Never copy.envfiles into it. - Never seed a checkout with a patch or share one checkout between agents.
- New work gets a new agent and worktree. Never re-brief a finished agent.
What ships with it
5 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 3d ago First seen · 100 lines · 116 tokens per session scan A dfae6015b8ee
herdr-agents is a skill published in the GitHub repository skrrt-sh/skills (2 stars, last pushed yesterday), licensed MIT. It adds 116 tokens to every session and 1,311 once invoked, about $0.0005 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 2026-10-22.
Other skills, from other repositories
commit-push-pr
Full git workflow - creates branch, commits, pushes, and creates or updates a PR with summary and test plan.
commit
Create a single well-crafted git commit from current changes. Analyzes diff, follows repo's commit style, and writes a concise "why not what" message.
fix-github-issue
Turn one known GitHub issue into a merged pull request, headless: a worker in its own git worktree, a draft pull request carrying its proof, an isolated second-engine reviewer, and a serial pull master that checks import-closure staleness against the base before it merges. Use when the task is "fix issue N…
changelog-assembly
This skill should be used when the user mentions "changelog", "release notes", "keep a changelog", "conventional commits", "commit message", "what changed", or "release summary". Provides a standardized methodology to turn raw, possibly non-conventional git history into human-readable changelogs and release notes.
git-for-research-code
When the user wants to version-control optimization research code - small commits per experiment change, tags for paper result snapshots, .gitignore for solver logs, linking result tables to commit hashes, and branch strategy for risky refactors. Also use when the user mentions "git workflow," "version control…
avi-pr-flow
Automate PR workflow with commit, push, create PR, and review cycle.