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/multiwindowgit clone --depth 1 https://github.com/dork-labs/dorkosWhat 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 | $0.00019 | $0.00688 |
| Opus 5 | $0.00010 | $0.00344 |
| Sonnet 5 | $0.00004 | $0.00138 |
| Haiku 4.5 | $0.00002 | $0.00069 |
Grade C, and why
multiwindow scanned grade C with 2 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Downloads and executes remote codehighSupply chain
curl | sh runs whatever the server returns today, which is not necessarily what it returned when this was reviewed.
curl -s "$BASE/api/config" | python3 -c "import sys,json;print('dorkHome:', json.load(sys.stdin)['dorkHome'])" Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
curl -sf "$BASE/api/health" | grep -q '"ok"' || echo "no server at $BASE — start one with 'pnpm dev' or 'pnpm dev:dogfood'" How it starts
The opening of the file, as written. The whole thing — 67 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Multi-window check
Open several cockpit windows against a running DorkOS, send a real message in
each, and report a PASS/FAIL table. Use it to go looking — for the regression
guard that runs in CI, see apps/e2e/tests/streams/multi-window.spec.ts.
Parse $ARGUMENTS
- A bare number →
--windows(default6) - Anything starting with
http→--base(defaulthttp://localhost:6241) --headed→ watch it happen
Preflight
An instance must already be running, with at least one agent. Confirm it and report the agent count before driving anything:
BASE="${BASE:-http://localhost:6241}"
curl -sf "$BASE/api/health" | grep -q '"ok"' || echo "no server at $BASE — start one with 'pnpm dev' or 'pnpm dev:dogfood'"
curl -s "$BASE/api/config" | python3 -c "import sys,json;print('dorkHome:', json.load(sys.stdin)['dorkHome'])"
If there is no server, stop and say so — do not start one silently, because the data directory and ports are the operator's choice.
Run
pnpm --filter @dorkos/e2e multi-window -- --windows <N> --base <URL>
It exits non-zero if any check fails.
Report
Give the PASS/FAIL table verbatim, then interpret it. What the failures mean:
| Failing check | Read it as |
|---|---|
| the app still answers with N windows open | the connection budget is exhausted — every other failure in the run is probably this one, so fix it before believing anything else |
| every window shows its own reply | streaming is broken under concurrency |
| no window shows another window's conversation | per-session state is leaking between windows |
| the browser matches the runtime transcript | the agent answered and the UI never showed it — check the stream, not the runtime |
| returning to an agent reopens the conversation you were having | session resolution regressed (DOR-928's shape) |
| each window showed the agent working | usually the list stream, which the connection budget starves first |
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.
- 2d ago First seen · 67 lines · 19 tokens per session scan C 2aea8194f566
multiwindow is a command published in the GitHub repository dork-labs/dorkos (9 stars, last pushed 2d ago), licensed MIT. It adds 19 tokens to every session and 688 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it C with 2 findings (downloads and executes remote code, 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
create-excalidraw-flowchart
Create a flowchart visualization in Excalidraw format for processes, pipelines, or logic flows.
create-excalidraw-dataflow
Create data flow diagrams (DFD) in Excalidraw format.
testarch-framework
Initialize production-ready test framework architecture (Playwright or Cypress) with fixtures, helpers, and configuration.
e2e
Generate and run end-to-end tests with Playwright. Creates test journeys, runs tests, captures screenshots/videos/traces.
axstream-teach
Compile a UI task into a fast replayable macro (do it once, instant forever).
capture_full_page
Capture a full-page (entire scrollable page, beyond the viewport) or single-element screenshot through the Chrome debugger and save it under screenshots/. Unlike takescreenshot (viewport only, no debugger), this uses CDP to render the whole page or a specific element.