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/dork-labs/dorkos/rooms-testgit clone --depth 1 https://github.com/dork-labs/dorkosWrote 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/dork-labs/dorkos/rooms-test)<a href="https://agentmods.dev/commands/dork-labs/dorkos/rooms-test"><img src="https://agentmods.dev/badge/commands/dork-labs/dorkos/rooms-test.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 | $0.00054 | $0.11561 |
| Opus 5 | $0.00027 | $0.05781 |
| Sonnet 5 | $0.00011 | $0.02312 |
| Haiku 4.5 | $0.00005 | $0.01156 |
Grade E, and why
rooms-test scanned grade E with 3 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 5d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Downloads and executes remote codehighSupply chain
curl | sh runs whatever the server returns today, which is not necessarily what it returned when this was reviewed.
curl -s "http://localhost:$API_PORT/api/mesh/agents" | python3 -c " Recursive force deletehighDestructive command
rm -rf with a variable or a broad path is one typo away from removing the wrong tree.
dotenv -- sh -c 'rm -rf /tmp/dorkos-test-mode-4243 && turbo run build --filter=@dorkos/server && pnpm --filter @dorkos/server exec tsx src/index.ts' Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
curl -sf -X PATCH http://localhost:4248/api/config -H 'content-type: application/json' \ How it starts
The opening of the file, as written. The whole thing — 534 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Self-test the parts of DorkOS that only exist when more than one participant shares a room: a channel with two agents and you, driven through a message burst, a mid-turn arrival, a halt, a reaction, a thread, and the three-way rule that keeps you in any room two agents share. The command drives the browser, cross-checks the room API and the SQLite rows underneath it, and writes an evidence-based report.
It is the rooms sibling of /chat:self-test (one session, in depth) and /chat:session-switch-test (two sessions, switching). Use it whenever you touch RoomService, the trigger dispatcher, the collector, room notices, reactions, threads, or the room event stream.
Every check below cites the capability row it verifies in meta/chat-capabilities.md — a run of this command is a report against that contract, not a vibe check.
Argument Parsing
Parse $ARGUMENTS:
- URL — any arg starting with
http. Defaulthttp://localhost:6241/channels(live) orhttp://localhost:4248/channels(sandbox — see Phase 1). mode:sandbox|live— which stack the run drives (meta/chat-capabilities.md§11).mode:sandbox— the test-mode runtime: throwaway data dir, deterministic, no model spend. Room turns really happen (thechromium-team-roome2e project proves an agent answers under this runtime), so every mechanical check below is reachable. What is not reachable is every check whose verdict is a judgment about what the agent chose to say — those are markedN/A (sandbox), never PASS.mode:live— the dev stack with a real runtime. Two agents answer per burst, so budget accordingly.- If the invocation does not state a mode, ASK the user before spending anything (
AskUserQuestion, offering both with their costs). Never assumelive.
perm:<mode>— permission mode for the agents' room-bound sessions. DefaultacceptEdits, which is the right default here and not a shortcut: a pending tool approval stalls a room turn, and a stalled turn makes checks 1–4 unreadable (you cannot tell "the agent folded three messages into one reply" from "the agent never got past a gate"). Usedefaultonly when the run is deliberately about approvals in rooms, and expect the downstream checks to come back BLOCKED.model:<id>— defaultclaude-haiku-4-5. Ignored undermode:sandbox.agents:a,b— the two agents to seat in the channel. Default: the first two agents the create dialog's agent search offers. Pick two whose display names are not prefixes of one another — half the assertions here locate an entry by its author's rendered name, andboinsidebobbywill find the wrong row.
Store as TEST_URL, MODE, PERM_MODE, MODEL, AGENT_A, AGENT_B.
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 · 534 lines · 54 tokens per session scan E 0a36fe7dcd93
rooms-test is a command published in the GitHub repository dork-labs/dorkos (9 stars, last pushed today), licensed MIT. It adds 54 tokens to every session and 11,561 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it E with 3 findings (downloads and executes remote code, recursive force delete, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other commands, from other repositories
dark-factory-behavioral-e2e
Holdout-pattern E2E validator. Drives agent-browser against the running DynaChat app to verify that the PR's user-facing behavior actually matches what the linked issue asked for.
testarch-framework
Initialize production-ready test framework architecture (Playwright or Cypress) with fixtures, helpers, and configuration.
auto-verify
프론트엔드 UX 검증 — Playwright 기반 비주얼 검증을 실행합니다.
record
Record a browser walkthrough of a URL using Antigravity (agy). Generates .webm video, screenshots, and a report. Auto-converts to MP4 if ffmpeg is available.
visual-verify
Use Playwright MCP to visually verify the UI that was just built or modified.
qa
Smoke or browser-walk a running app. Report only. Do not implement. Do not merge.