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 agentmods add instructions/zaxbyhub/opencode-swarm/claude-mdgit clone --depth 1 https://github.com/ZaxbyHub/opencode-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/instructions/zaxbyhub/opencode-swarm/claude-md)<a href="https://agentmods.dev/instructions/zaxbyhub/opencode-swarm/claude-md"><img src="https://agentmods.dev/badge/instructions/zaxbyhub/opencode-swarm/claude-md.svg" alt="Measured on agentmods" 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 | $0.01340 | $0.01340 |
| Opus 5 | $0.00670 | $0.00670 |
| Sonnet 5 | $0.00268 | $0.00268 |
| Haiku 4.5 | $0.00134 | $0.00134 |
Grade A, and why
opencode-swarm CLAUDE.md 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 yesterday.
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 — 97 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Claude Code Swarm Mode
For this repository, AGENTS.md is the root engineering contract. Read it before any code change. The long-form rationale and historical failure map live in
docs/engineering-invariants.md. When swarm mode is enabled,AGENTS.mdstill applies; swarm-mode instructions add workflow structure, not exceptions to the engineering invariants.
Normal behavior is the default.
If .claude/session/swarm-mode.md exists, swarm mode is enabled for the current session and you must read that file before starting complex work.
When swarm mode is enabled:
- Quality is the only success metric.
- There is no time pressure.
- Do not compress a workflow just because the task is large.
- Prefer parallel subagents for disjoint investigation and review work.
- Keep implementation, validation, and final judgment in separate contexts when possible.
- Explorer-style work is for breadth and candidate generation.
- Reviewer-style work is for validation of candidate findings or implementation quality.
- Critic-style work is for final challenge of reviewer-confirmed findings or high-impact implementation conclusions.
- Do not let the same context both invent and approve a finding when a separate verification pass is possible.
- No approval without positive evidence of what was checked.
- No high-severity finding without exact evidence and, when relevant, runtime-aware verification.
- Preserve Claude Code speed by parallelizing broadly and reserving the deepest validation for high-risk or ambiguous work.
- Across many different repositories, explore local patterns first rather than assuming one project's conventions apply to another.
If .claude/session/swarm-mode.md does not exist, behave normally.
Permanent project directives (do not override)
These directives were explicitly requested by the project owner and persist across all sessions, contributors, and agents. They are not workflow preferences — they are correctness gates.
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.
- yesterday First seen · 97 lines · 1,340 tokens per session scan A a67070bc7f79
opencode-swarm CLAUDE.md is an instructions file published in the GitHub repository ZaxbyHub/opencode-swarm (464 stars, last pushed today), licensed MIT. It adds 1,340 tokens to every session, about $0.0067 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 2026-09-03.
Other instructions, from other repositories
dev-3.0 AGENTS.md
AGENTS.md instructions for h0x91b/dev-3.0, covering agents.md, response style, what is this, ui/ux work — which skill, and when and where dev3 overrides the better-family skills.
meridian CLAUDE.md
Claude Code instructions for rynfar/meridian, covering claude.md, what this is, commands, code rules and module boundaries.
meridian AGENTS.md
AGENTS.md instructions for rynfar/meridian, covering agents.md, what this is, commands, code rules and module boundaries.
stella AGENTS.md
AGENTS.md instructions for macanderson/stella, covering agents.md, essential commands, the gate — what every push is held to, architecture: ports, not direct dependencies and the definition of done: witness tests.
stella CLAUDE.md
Claude Code instructions for macanderson/stella, covering claude.md and hard rules for every session.
learn-likecc AGENTS.md
AGENTS.md instructions for Harzva/learn-likecc, a project described as: Claude Code 源码逆向恢复项目 | Source Map 逆向 · 架构分析 · 可运行版本 | 1900+ 文件 · 51万行代码 · 12章节课程.