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/siddsachar/row-bot/computer_use_guidenpx skills add siddsachar/row-bot --skill computer_use_guidegit clone --depth 1 https://github.com/siddsachar/row-botWhat 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 | $0.00004 | $0.00490 |
| Opus 5 | $0.00002 | $0.00245 |
| Sonnet 5 | $0.00001 | $0.00098 |
| Haiku 4.5 | $0.00000 | $0.00049 |
Grade A, and why
computer_use_guide 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 3d 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.
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.
What it actually says
COMPUTER USE WORKFLOW
- Prefer Browser for managed pages and Computer Use for native targets; never switch silently. Issue one Computer Use call at a time. Use one app-scoped capture with an exact current
list_appsname, never parallel app/window discovery. target_idand tokens live in the current Computer Use generation. A new user turn starts with discovery or app-scoped capture. Say gone or its lease expired only after a previously returned opaque target producestarget_gone.- Do not repeat after
app_not_running,window_not_found, protected refusal, or non-retryable native capture. Exact-filter ambiguity permits a returned current token without another capture only for the present observation/lease. - On stale refusal, capture the exact same target once and retry the same exact action once. If stale repeats, Stop or request Take over. Do not switch action family or delivery engine.
- Token-bound
typerejects disabled, read-only, secure, protected, or structural targets. Tokenless type is literal text insertion; usereplace_textfor a complete value. - Separate action dispatched, native state change, and exact postcondition. Read
driver_verdict,degraded,escalation_recommendation,verdict, andnext_stepliterally:done,verify_fresh_state,escalate, ortake_overapplies only to this action receipt. - Start with
delivery_mode=auto. Use explicitdelivery_mode=foregroundonly for click, double-click, right-click, type, key, or scroll after fresh exact-target evidence and a prior structured recommendation/refusal. Pre-dispatchbackground_unavailableorforeground_requiredpermits one same-action foreground rung; a second refusal requires Take over. Never focus first or assume focus persists. - Text insertion, accepted or uncertain mutations, timeouts, disconnects, and ended sessions must not be replayed. Never replay them. After one fresh unchanged
suspected_noop, a reversible click permits one alternative exact route: recommendedforeground, orpxfrom a new screenshot-grounded exact-target capture.pageis observable but unsupported; never use SOM, blind coordinates, Browser/CDP/Page, shell, clipboard, or app-specific fallbacks.
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.
- 3d ago First seen · 18 lines · 4 tokens per session scan A 6f95afecb10c
computer_use_guide is a skill published in the GitHub repository siddsachar/row-bot (1,479 stars, last pushed 5d ago), licensed Apache-2.0. It adds 4 tokens to every session and 490 once invoked, about $0.0000 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
agent-browser
Браузерная автоматизация для интерактивных веб-задач. Используй, когда нужно открыть сайт, заполнить форму, кликнуть, сделать скриншот, залогиниться, спарсить JS-страницу или протестировать веб-приложение. Триггеры — «зайди на сайт», «заполни форму», «нажми кнопку», «скриншот страницы», «войди в аккаунт», «спарси…
test-ui
Test the Wayfinder agent extension UI by starting the dev environment and visually verifying changes via CDP. Covers the new tab page (left sidebar — Home, Scheduled Tasks, Settings, etc.) and the right side panel (chat interface). Use after making UI changes to apps/agent/.
webapp-testing
Toolkit for interacting with and testing local web applications using Playwright. Supports verifying frontend functionality, debugging UI behavior, capturing browser screenshots, and viewing browser logs.
add-tavily-tool
Add Tavily Search and Extract as keyless remote MCP tools for selected NanoClaw agent groups. Use when installing Tavily web search or URL extraction without an API key.
agent-browser
Browse the web for any task — research topics, read articles, interact with web apps, fill forms, take screenshots, extract data, and test web pages. Use whenever a browser would be useful, not just when the user explicitly asks.
playwright-pom-discovery
Use when building or extending a Page Object Model (POM) for the Opik E2E suite (under testsendtoend/e2e/pom/) and you need to choose stable selectors against the live UI. Walks through seeding required state, exploring the running page with the Playwright MCP (accessibility snapshot + data-testid enumeration)…