CWC Workshops is a collection of materials from Anthropic-run workshops on building and evaluating AI-assisted coding workflows. The workshops cover model selection, multi-agent systems, managed agents, and product development with coding agents. The catalogue entries are examples and teaching materials from those workflows.
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 commands/anthropics/cwc-workshops/cwc-setupgit clone --depth 1 https://github.com/anthropics/cwc-workshopsWrote 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/commands/anthropics/cwc-workshops/cwc-setup)<a href="https://agentmods.dev/commands/anthropics/cwc-workshops/cwc-setup"><img src="https://agentmods.dev/badge/commands/anthropics/cwc-workshops/cwc-setup.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 5.1 | $0.00023 | $0.01179 |
| Opus 5 | $0.00012 | $0.00589 |
| Sonnet 5 | $0.00005 | $0.00236 |
| Haiku 4.5 | $0.00002 | $0.00118 |
Grade D, and why
cwc-setup scanned grade D with 3 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 6d 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.
Asks for rootmediumPrivilege escalation
A mod that escalates privileges can change anything on the machine, not only the project.
| `java not found` | macOS: `brew install openjdk@21 && sudo ln -sfn $(brew --prefix openjdk@21)/libexec/openjdk.jdk /Library/Java/JavaVirtualMachines/`. Linux: `sudo apt install openjdk-21-jdk`. Then re-run. | Recursive force deletehighDestructive command
rm -rf with a variable or a broad path is one typo away from removing the wrong tree.
| `npm install (bot) failed` | Likely a corp registry. Run `(cd bot && rm -rf node_modules package-lock.json && npm install --registry=https://registry.npmjs.org)` then re-run `./setup.sh`. | Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
| `bot did not register with the relay` | Read `/tmp/mc-bot.log`, look for `[relay]` lines. Check `curl $EVENT_URL/api/config` works from their machine. If the event server is fine but WebSockets can't connect, the venue How it starts
The opening of the file, as written. The whole thing — 82 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are setting up the Agent Battle workshop stack for a participant.
Your job: get them from a fresh clone to ▸ ready with no manual
debugging on their part. Be adaptive — every machine is a little
different.
Steps
1. Check the participant env vars that must be exported:
for v in PARTICIPANT MINECRAFT_EULA; do
if [ -z "${!v:-}" ]; then echo "MISSING: $v"; else echo "OK: $v"; fi
done
[ -f .env.event ] && grep EVENT_URL .env.event || echo "NO .env.event"
If any are MISSING, stop and tell them what to export:
ANTHROPIC_API_KEY— their own key from console.anthropic.com (optional if their environment has SDK OAuth/workload-identity auth)PARTICIPANT— any unique name they pickMINECRAFT_EULA— must beaccept. Show them https://www.minecraft.net/eula and ask whether they agree. If yes, they runexport MINECRAFT_EULA=accept. You must not set this for them or assume agreement — the user has to affirmatively accept the EULA themselves.
EVENT_URL, LEADERBOARD_KEY, and MC_SEED come from the
.env.event file in this directory — setup.sh reads it
automatically. Do NOT block on these. (If .env.event has an empty
EVENT_URL, setup runs in SOLO mode — local event server + one
quick-tunnel — which is correct for at-home practice.)
Do NOT proceed until PARTICIPANT and MINECRAFT_EULA are set (ANTHROPIC_API_KEY is optional).
2. Run setup and capture the full output:
./setup.sh 2>&1
3. Decide based on what you see:
If the last line is the boxed OPEN THIS IN YOUR BROWSER block →
success. Tell them:
- Open
http://localhost:8088/viewin a browser (their bot's camera) - Open
my_agent.pyand find theAGENT = dict(...)block — that's what they'll edit - Run
python3 my_agent.py— every run is 5 min and posts to the leaderboard; best run counts, so iterate freely python3 my_agent.py --evalfor a ~30-60s scorecard with no run- You're standing by if anything breaks — they can type
/cwc-fix
If you see ✗ (a die() line) → failure. Match the error to the
playbook in CLAUDE.md and remediate. Common cases:
| Error pattern | Fix |
|---|---|
java not found |
macOS: brew install openjdk@21 && sudo ln -sfn $(brew --prefix openjdk@21)/libexec/openjdk.jdk /Library/Java/JavaVirtualMachines/. Linux: sudo apt install openjdk-21-jdk. Then re-run. |
npm install (bot) failed |
Likely a corp registry. Run (cd bot && rm -rf node_modules package-lock.json && npm install --registry=https://registry.npmjs.org) then re-run ./setup.sh. |
server failed — see /tmp/mc-server.log |
Read that log. If Address already in use: stale java holding :25565 → see CLAUDE.md "Killing leftovers". If Unsupported class file: Java too old, install 17+. |
bot failed — see /tmp/mc-bot.log |
Read that log. Usually Cannot find module (npm install incomplete) or ECONNREFUSED (server not up yet — wait 10s, re-run). |
bot did not register with the relay |
Read /tmp/mc-bot.log, look for [relay] lines. Check curl $EVENT_URL/api/config works from their machine. If the event server is fine but WebSockets can't connect, the venue network may block them — escalate to the facilitator. |
event server failed (SOLO mode) |
Read /tmp/event-local.log. Usually a missing (cd event && npm install). |
tunnel failed (SOLO mode only) |
Transient cloudflared issue. Re-run ./setup.sh. If /tmp/cloudflared is missing or zero-byte, delete it so tunnel.sh re-downloads. |
ANTHROPIC_API_KEY not set warning |
Fine to proceed if their environment has SDK auth; otherwise have them export it. |
After remediating, re-run ./setup.sh (it's idempotent). Loop until
▸ ready or you hit something not in the playbook — then explain what
you see and ask if they want you to dig in.
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.
- 6d ago First seen · 82 lines · 23 tokens per session scan D 83d6a4629412
cwc-setup is a command published in the GitHub repository anthropics/cwc-workshops (2,045 stars, last pushed 9d ago), licensed Apache-2.0. It adds 23 tokens to every session and 1,179 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it D with 3 findings (asks for root, recursive force delete, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other commands, from other repositories
checklist
Generate a custom checklist for the current feature based on user requirements.
clarify
Identify underspecified areas in the current feature spec by asking up to 5 highly targeted clarification questions and encoding answers back into the spec.
specify
Create or update the feature specification from a natural language feature description.
analyze
Perform a non-destructive cross-artifact consistency and quality analysis across spec.md, plan.md, and tasks.md after task generation.
converge
Assess the current codebase against the feature's spec, plan, and tasks, then append any remaining unbuilt work as new tasks to tasks.md so implement can complete it.
implement
Execute the implementation plan by processing and executing all tasks defined in tasks.md.