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/copilotkit/openbot/openbot-data-accessnpx skills add CopilotKit/OpenBot --skill openbot-data-accessgit clone --depth 1 https://github.com/CopilotKit/OpenBotWhat 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 | $0.00189 | $0.04603 |
| Opus 5 | $0.00095 | $0.02302 |
| Sonnet 5 | $0.00038 | $0.00921 |
| Haiku 4.5 | $0.00019 | $0.00460 |
Grade A, and why
openbot-data-access 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 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.
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 — 293 lines — stays where its author put it; the contents beside it link to each section on GitHub.
OpenBot Data Access
When To Use
This skill applies to any change under app/src that moves data between the browser and the API
server. It fires on new screens, new endpoints, new query keys, and on any diff that introduces
fetch anywhere but app/src/lib/client.ts.
It does not cover server handlers under server/, zod form schemas (lib/<entity>/form.ts), or page
layout. It does cover lib/copilot/: the conversation itself streams over AG-UI, but the tool calls a
Bot makes during a turn are ordinary authenticated requests and go through the client like everything
else.
The Shape
Every entity the browser knows about owns a directory under app/src/lib/:
app/src/lib/
client.ts # the only fetch in the app
<entity>/
queries.ts # read types, key factory, queryOptions factories
mutations.ts # input type, mutationOptions factories
form.ts # zod schema (a different skill's territory)
client.ts owns the transport: credentials, the JSON content type, body serialisation, and turning a
failed status into an Error carrying the server's own message. It owns nothing about meaning — the
envelope key and the sentence a person reads stay at the call site, because those are facts about one
endpoint rather than about requests in general.
client<T>(path, key, options?): Promise<T> // parsed, and `key` unwrapped
client(path, options?): Promise<Response> // for a caller that only needed it to work
tryClient(path, options?): Promise<Response> // never throws; the status is the answer
options is { method?, body?, fallback?, signal? }. body is serialised by the client, which is
also what sets the content type — so a caller passes an object, never a string. Passing
JSON.stringify(x) sends a JSON string of a JSON string, which no endpoint accepts.
Three kinds of request
Not everything crossing the wire is cached state, and the shape follows from which kind it is.
- A cached read is a
queryOptionsfactory inqueries.ts. It has a key, and something can invalidate it. - A write somebody asked for is a
mutationOptionsfactory inmutations.ts. It invalidates on success. - Everything else is a plain exported function, living beside the factories for its entity.
A verdict about this moment (
decideComponent,testAgentConnection), a tool call during a Bot's turn (callPluginTool, the computer control surface), a frame of a screen, a step inside another write (storeMcpToken). These fail closed and return a value rather than throwing, because a refusal is usually the answer. Giving one a cache key would create a key nothing reads and an invalidation nothing triggers.
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 · 293 lines · 189 tokens per session scan A 39aac85348d6
openbot-data-access is a skill published in the GitHub repository CopilotKit/OpenBot (3,511 stars, last pushed 4d ago), licensed MIT. It adds 189 tokens to every session and 4,603 once invoked, about $0.0009 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
opencli-usage
Use at the start of any OpenCLI session — this is the top-level map of what opencli can do, how to discover adapters, what flags and output formats are universal, and which specialized skill to load next. Point here when an agent asks "what can opencli do?" or "how do I find the right command?".
opencli-sitemap-author
Use when creating or maintaining OpenCLI site sitemaps: agent-facing navigation, page-state, action, workflow, API-reference, pitfall, and fallback knowledge for a website. Use after browser exploration discovers durable site context, when a sitemap is stale, or when promoting local site knowledge into the repo.
opencli-browser-sitemap
Use when driving a website with opencli browser and sitemap context is available, requested, or needed to avoid blind navigation. Guides agents to consume site sitemap files lazily, choose adapter/browser fallback paths, resume from state signatures, and mark stale sitemap entries without trusting them over live…
pinchtab-mcp
Use this skill when a task requires browser automation through PinchTab's MCP server connected to a remote browser instance. Covers navigation, element interaction, data extraction, form filling, multi-step flows, and session management via MCP tools.
pinchtab-stealth-score
Run the PinchTab stealth-score sweep against 15 bot-detection / fingerprint sites (sannysoft, rebrowser, deviceandbrowserinfo, iphey, whoer, browserscan, pixelscan, fingerprint-scan, incolumitas, fvision, amiunique, browserleaks, creepjs, coveryourtracks, fingerprint-demo). Starts a Docker PinchTab container per…
webcmd-usage
Use at the start of any Webcmd session. This is the top-level map of what webcmd can do, how to discover adapters, what flags and output formats are universal, and which specialized skill to load next. Point here when an agent asks "what can webcmd do?" or "how do I find the right command?".