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/untrivial-ai/agent-orchestrator/ao-desktop-devnpx skills add Untrivial-ai/agent-orchestrator --skill ao-desktop-devgit clone --depth 1 https://github.com/Untrivial-ai/agent-orchestratorWhat 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 | $0.00088 | $0.01827 |
| Opus 5 | $0.00044 | $0.00914 |
| Sonnet 5 | $0.00018 | $0.00365 |
| Haiku 4.5 | $0.00009 | $0.00183 |
Grade A, and why
ao-desktop-dev 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 — 144 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AO Desktop Dev
Run the Electron application from the current checkout and verify it in the native window. Keep the dev daemon, renderer, Electron preload, and selected data mode explicit so the user never reviews a mock or stale build by accident.
Choose the data mode
Ask only when the request does not make the desired data source clear.
- Use isolated mode by default for implementation and destructive testing. Electron uses port
3002,~/.ao/dev/running.json,~/.ao/dev/data, and~/.ao/dev/electron. - Use real-data mode only when the user explicitly asks to see this machine's actual AO projects or sessions. Start the checkout's dev daemon on the isolated dev port/run file while pointing
AO_DATA_DIRat the real AO data directory. This is a separate daemon process using real data; do not describe it as the installed app's daemon. - Never try to attach an unpackaged Electron app directly to a packaged daemon from another checkout. The supervisor intentionally rejects daemon identity mismatches.
- Warn before actions in real-data mode that create, terminate, rename, or otherwise mutate sessions. Merely opening and inspecting the UI is expected.
Preflight
- Work from the repository root and inspect
git status --short --branch. - Read
frontend/package.json,frontend/src/main.ts, andfrontend/forge.config.tswhen the launch behavior may have changed. Treat source as authoritative over this skill. - Confirm Node/npm and the Go version required by
backend/go.modare available. Runnpm ciinfrontend/only when dependencies are absent or inconsistent; do not reinstall on every launch. - Check for an existing dev instance from this exact checkout before starting another. Do not kill by process name alone, and never terminate every Electron process.
On macOS/Linux, inspect checkout-scoped processes and listeners with finite commands:
ps -axo pid=,ppid=,pgid=,lstart=,command= | rg 'frontend/(node_modules/.bin/electron-forge|node_modules/electron/dist/Electron)'
lsof -nP -iTCP:3002 -sTCP:LISTEN
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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 · 144 lines · 88 tokens per session scan A a549d9f6362b
ao-desktop-dev is a skill published in the GitHub repository Untrivial-ai/agent-orchestrator (10,763 stars, last pushed today), licensed Apache-2.0. It adds 88 tokens to every session and 1,827 once invoked, about $0.0004 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 2026-08-30.
Other skills, from other repositories
ai-product-photography
Generate professional AI product photography and commercial images. Models: FLUX, Imagen 3, Grok, Seedream for product shots, lifestyle images, mockups. Capabilities: studio lighting, lifestyle scenes, packaging, e-commerce photos. Use for: e-commerce, Amazon listings, Shopify, marketing, advertising, mockups.…
contributor-onboarding
Help a new contributor get productive on this checkout - inspect sync state against main, build, run the repository's exact verification gate, and produce a local what's-new digest. Never fetches, pulls, or modifies a dirty tree on its own. Explicit-only.
bernstein-run
Run a verified multi-agent goal with Bernstein. Use when a task is too large for a single agent session: Bernstein decomposes the goal into tasks, spawns CLI coding agents in parallel git worktrees, verifies their output, and merges results. Also use to check run status, costs, and to verify a finished run against its…
bernstein-plan
Create and manage multi-step execution plans in Bernstein. Plans decompose complex goals into stages with dependencies. Use when the user wants to plan a complex feature, break down a large task, or review an execution plan before agents start working.
bernstein-approve
Review and approve/reject pending tasks or plans in Bernstein. Use when the user asks about approvals, wants to review agent work, or needs to approve/reject a plan before execution begins.
bernstein-quality
Show quality metrics for Bernstein runs - success rates per model, lint/test pass rates, completion time distributions. Use when the user asks about quality, reliability, which model performs best, or pass rates.