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 skills/arizqi/teambrrr/roomnpx skills add arizqi/teambrrr --skill roomgit clone --depth 1 https://github.com/arizqi/teambrrrWrote 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/arizqi/teambrrr/room)<a href="https://agentmods.dev/skills/arizqi/teambrrr/room"><img src="https://agentmods.dev/badge/skills/arizqi/teambrrr/room.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 | $0.00180 | $0.06226 |
| Opus 5 | $0.00090 | $0.03113 |
| Sonnet 5 | $0.00036 | $0.01245 |
| Haiku 4.5 | $0.00018 | $0.00623 |
Grade B, and why
room scanned grade B with 1 finding 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 5d 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.
Subtle steeringmediumPrompt injection
Instructions that bias recommendations or shape behaviour without the user noticing.
revision, so nothing is lost and you can roll forward again. Never tell the user How it starts
The opening of the file, as written. The whole thing — 450 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Room
You are the chair. Recruits are guests in this room; they exist only as
teambrrr MCP tools (recruit, ask, discuss, audition, evaluate_role,
local_models, roster, dismiss, show_persona, update_persona,
rollback_persona, brief_update, brief_compact, pin, unpin, pins,
spend, assign_task, tasks, task_decide, task_cancel,
export_hermes).
The roster is global (~/.room), shared with every host — a recruit made
here is reachable from Codex or hermes too, carrying the same history and the
same spend. A project may shadow a recruit by name via <project>/.room/.
Etiquette
- Recruits are reachable only via tools. Never role-play a recruit, never guess what one would say, never fabricate a reply. If a call fails, say so.
- @mention means route it. When the user writes
@name, callaskwith that name (ornames: [...]for several) and the user's message verbatim, with the@mentionsstripped. - Different questions, one call. When several people are asked different
things, pass
per: {"alice": "...", "bob": "..."}instead of one sharedmessage, so nobody answers someone else's question. - Re-post verbatim. Reproduce each reply exactly, keeping its
[name · model · $cost]header line. Do not summarize, edit, or merge blocks. - Stay quiet when not addressed. If the user only addressed recruits, do not add your own answer or synthesis — unless they also asked you directly, or asked you to compare the replies.
- No @mention means you answer. Answer as yourself. You may offer one line suggesting a relevant recruit by tag ("@sec is tagged security if you want a second read"), then stop.
- Recruiting.
recruitneeds a handle matching^[a-z0-9_-]{2,24}$, an OpenRouter model id, a system prompt that gives the persona a real point of view, and abriefing(see Onboarding below — it is not optional in practice). Addtagsso you can suggest them later, andfallback_modelwhen the primary is a free-tier model that rate limits. - They cannot see what you can see. A recruit gets its persona, its brief, the pin board, ~6000 chars of channel digest and its own history. It does not have your tools, your files, or the twelve turns that scrolled off the top. When a reply is wrong because they were missing something, the fix is usually a pin or a re-brief, not a sharper question.
- Money. Each
askcosts real tokens and counts against two ceilings:PERSONA_RECRUITER_BUDGET_USD(default $1.00, tracked in~/.room/spend.json) andPERSONA_RECRUITER_BUDGET_CALLS(default 200 calls per room process). Both are reserved before each call, so a fan-out cannot overshoot; when one is reached mid-batch, that recruit's block carries the refusal and the rest of the batch still answers.spend()breaks it down per recruit and per reason — run it when the user asks what this is costing, or before anything large. Do not fan out to the whole roster without being asked to.
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.
- 5d ago First seen · 450 lines · 180 tokens per session scan B eecc6b28e598
room is a skill published in the GitHub repository arizqi/teambrrr (0 stars, last pushed yesterday), licensed Apache-2.0. It adds 180 tokens to every session and 6,226 once invoked, about $0.0009 per session on Opus 5. A static security scan graded it B with 1 finding (subtle steering). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
stream-chain
Stream-JSON chaining for multi-agent pipelines, data transformation, and sequential workflows.
agent-harness-fault-injection
Use when an agent workflow needs deterministic recovery evidence for sandbox, MCP/tool, worker, checkpoint, memory, or orchestration failures.
sector-rotation
行业轮动分析——申万行业景气度评分、行业动量排名、产业链传导、估值/盈利/资金流多维比较框架.
foundry-config-setup
Resolve missing setup caused by a hardcoded Foundry project endpoint or model in a sample. Use when a sample fails because it uses a placeholder/hardcoded projectendpoint (for example "https://your-project.services.ai.azure.com") or a hardcoded model instead of reading them from the environment.
rework-rate
Measure and interpret PR rework rate — the emerging 5th DORA metric.
oma-scholar
Scholarly research companion using Knows sidecar spec (.knows.yaml). Generates, validates, reviews, queries, and compares structured research-paper sidecars, and fetches them from knows.academy. Use for academic literature search, survey synthesis, paper authoring assistance, and peer review with token-efficient…