Getting it into your agent
It runs from inside its repository, so the clone comes first — what it calls does not travel with the file alone.
git clone --depth 1 https://github.com/richfrem/agent-plugins-skillsnpx agentmods add skills/richfrem/agent-plugins-skills/agent-swarmWrote 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/richfrem/agent-plugins-skills/agent-swarm)<a href="https://agentmods.dev/skills/richfrem/agent-plugins-skills/agent-swarm"><img src="https://agentmods.dev/badge/skills/richfrem/agent-plugins-skills/agent-swarm.svg" alt="Measured on agentmods" 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 | $0.00074 | $0.02071 |
| Opus 5 | $0.00037 | $0.01035 |
| Sonnet 5 | $0.00015 | $0.00414 |
| Haiku 4.5 | $0.00007 | $0.00207 |
Grade A, and why
agent-swarm 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 4d 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 — 163 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Dependencies
This skill requires Python 3.8+ and standard library only. No external packages needed.
To install this skill's dependencies:
pip-compile ./requirements.in
pip install -r ./requirements.txt
See ../../requirements.txt for the dependency lockfile (currently empty — standard library only).
Agent Swarm
Parallel or pipelined execution across multiple agents and worktrees. The orchestrator partitions work, dispatches to agents, and verifies/merges the results.
When to Use
- Large features that can be split into independent work packages
- Bulk operations (tests, docs, migrations, RLM distillation) that benefit from parallelism
- Multi-concern work where specialists handle different aspects simultaneously
Process Flow
- Plan & Partition -- Break work into independent tasks. Define boundaries clearly.
- Route -- Decide execution mode:
- Sequential Pipeline -- Tasks depend on each other (A -> B -> C)
- Parallel Swarm -- Tasks are independent (A | B | C) 2.5. Interactively Determine CLI and Model (ask once during bootstrap): Before dispatching the swarm workers, you must ask the user:
- "Which LLM CLI engine would you like to run the swarm workers through?" (Options:
agy,claude,copilot,gemini,llama). - "Which specific model should be used?" (Options/defaults per engine, e.g.,
Gemini 3.5 Flash (Low)orgemini-3.5-flashforagy). - Construct the
swarm_run.pyinvocation with--engineand--modelmatching their choices, appending< /dev/nullto prevent TTY input halts (SIGTTIN).
- Dispatch -- Create a worktree per task. Assign each to an agent:
- CLI agent (Claude, Gemini, Copilot, Antigravity) using the selected setup
- Deterministic script
- Human
- Execute -- Each agent works in isolation. No cross-worktree communication.
- Verify & Merge (Trust But Verify & TDD) -- Orchestrator checks each worktree's output against acceptance criteria. No blind trust is allowed.
- TDD Enforcement: Prioritize running unit and integration tests to ensure no regressions were introduced.
- Delta Inspection: Check modified files directly for stubs, stales, or placeholders.
- Verify Quality: If verification fails, generate a correction packet, reject, and re-dispatch.
- Pass -> Merge into main branch
- Seal -- Bundle all merged artifacts
- Retrospective -- Did the partition strategy work? Was parallelism effective?
What ships with it
11 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.
- acceptance-criteria.md 775 B
- assets/resources/agent_swarm.mmd 43 B
- evals/evals.json 733 B
- evals/results.tsv 301 B
- fallback-tree.md 1.6 KB
- references/acceptance-criteria.md 42 B
- references/cheapest_models.json 40 B
- references/cheapest_models.md 38 B
- references/fallback-tree.md 36 B
- requirements.txt 22 B
- scripts/swarm_run.py 29 B runs code
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.
- 4d ago Changed 31f17b962959
- 8d ago First seen · 163 lines · 74 tokens per session scan A f6fb359441c2
agent-swarm is a skill published in the GitHub repository richfrem/agent-plugins-skills (6 stars, last pushed today), licensed MIT. It adds 74 tokens to every session and 2,071 once invoked, about $0.0004 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 2026-08-31.
Other skills, from other repositories
composio
Use Composio from Agent Swarm through the agent-swarm x composio CLI route, the swarmx MCP tool, or a registered ctx.api.composio script connection. Trigger when a task needs connected third-party app tools such as Gmail, Google Calendar, Google Docs, Google Drive, GitHub, Slack, Notion, or HubSpot through Tool Router…
close-issue
Close a GitHub or GitLab issue with a summary comment.
soul
Embody this digital identity. Read SOUL.md first, then STYLE.md, then examples/. Become the person—opinions, voice, worldview.
implement-issue
Implement a GitHub issue or GitLab issue and create a PR/MR.
bbc-skill
A read-only tool for collecting all comments from a Bilibili video, including replies and pinned comments. Bilibili is a Chinese video-sharing platform.
feature-implementation
Implement software features, fix bugs, and write production-quality code. Use this for hands-on development work including writing new code, modifying existing code, and debugging issues.