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 instructions/mozilla-ocho/tabstack-cli/claude-mdgit clone --depth 1 https://github.com/Mozilla-Ocho/tabstack-cliWhat 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 | $0.04305 | $0.04305 |
| Opus 5 | $0.02152 | $0.02152 |
| Sonnet 5 | $0.00861 | $0.00861 |
| Haiku 4.5 | $0.00430 | $0.00430 |
Grade B, and why
tabstack-cli CLAUDE.md scanned grade B 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.
Asks for rootmediumPrivilege escalation
A mod that escalates privileges can change anything on the machine, not only the project.
All reads and writes go through `config.CredentialStore` (`Load`/`Save`/`Path`) so an OS keychain implementation can be added without touching command code; `FileStore` is the only implementation. Saves are atomic (temp Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
`readInput`/`readJSON` (helpers.go) accept a literal string, `@file`, or `-` for stdin, mirroring curl's `-d`. `readJSON` validates JSON locally so a malformed schema fails with a clear message instead of an opaque API 4 How it starts
The opening of the file, as written. The whole thing — 121 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md
This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
Overview
tabstack is a Go CLI client for the Tabstack AI API: browser automation, web research, and structured extraction/generation from any URL. It is a hand-written HTTP client (no generated code) wrapped in a Cobra command tree.
Commands
make build # build into ./bin/tabstack (version stamped from `git describe`)
make install # install into $GOPATH/bin
make run ARGS="extract markdown --url ..."
make lint # gofmt -w . then go vet ./...
make test # go test ./...
make help # list all targets
Version is injected via -ldflags -X .../cmd.version=$(VERSION); VERSION comes from git describe --tags --always --dirty, falling back to dev.
Tests live alongside each package (*_test.go); GitHub Actions runs gofmt/vet/build/test on push and PRs (.github/workflows/ci.yml). client.WithHTTPClient injects a mock *http.Client for tests. scripts/smoke-test.sh (make smoke) exercises every command against the live API.
Architecture
Three layers, each in its own package:
cmd/tabstack/main.go: entry point; builds the root command and runs it throughfang.Execute(charmbracelet/fang), which renders styled help, errors, version output, and addscompletion/mansubcommands. fang prints errors itself and returns them, so main keeps owning the exit-code mapping: it checks for any error implementingCode() int(thecodedinterface) and exits with that code, else maps cobra usage errors to 2, else 1. Version is handed to fang viacmd.Version()(stamped by-ldflags). (Lives in its own dir sogo install .../cmd/tabstackproduces atabstackbinary.)cmd/: Cobra command tree, one file per endpoint group (agent,extract,generate,schema,auth,keys). Each leaf command only builds its request and calls the client. The auth flow is split acrossauth.go(status/logout/sessions),login.go(the OAuth flow and loopback callback),switch.go(org switching),keysetup.go(the shared key-setup prompt), andorgs.go(org selector resolution).internal/client/: the HTTP client and per-endpoint request/response types for the product host. Sends the org-scoped API key, never the session.internal/console/: the client for the auth and management host (console.tabstack.ai):/oauth/*(PKCE authorize URL, code exchange, refresh) and/cli/*(me, organizations, api_keys, sessions, logout). Sends the user-scoped session token, never an API key. Split by host on purpose so the wrong credential cannot reach the wrong endpoint.internal/schemas/: fetches the publictabstack-schemasGitHub repo and manages the local schema store (schema pull/schema list). Kept separate fromclient: it talks toraw.githubusercontent.comunauthenticated, so it must never see the API bearer token. Mirrorsclient's injectable-http.Clientpattern (schemas.WithHTTPClient) for tests.internal/config/: credential and base-URL resolution. Also resolvesSchemasDir()(the defaultschema pulldestination, alongside the config file under the tabstack config home).internal/ui/: output rendering (pretty vs JSON), styles, spinner.
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 · 121 lines · 4,305 tokens per session scan B c2c52244bae3
tabstack-cli CLAUDE.md is an instructions file published in the GitHub repository Mozilla-Ocho/tabstack-cli (11 stars, last pushed 26d ago), licensed MIT. It adds 4,305 tokens to every session, about $0.0215 per session on Opus 5. A static security scan graded it B with 2 findings (asks for root, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other instructions, from other repositories
codex AGENTS.md
AGENTS.md instructions for openai/codex, covering rust/codex-rs, the codex-core crate, code review rules, crate api surface and model visible context.
buildNext
Working notes and architecture documentation for the new esbuild-based build system in build/next. Use when making changes to the new build pipeline (transpile/bundle commands, NLS plugin, source-map handling, resource copying, or self-hosting watch tasks).
next.js AGENTS.md
Instructions for vercel/next.js, covering next.js development guide, codebase structure, monorepo overview, core package: packages/next and other important packages.
vscode oss-third-party-notices.instructions.md
Instructions for microsoft/vscode, covering vs code oss third-party-notices pipeline, architecture, pipeline flow in ci, applying the notice (cutover) and fallback chain (never fail the build).
spec-kit AGENTS.md
Instructions for github/spec-kit, covering agents.md, about spec kit and specify, quickstart — add a new integration in 5 steps, integration architecture and integrationmanifest — file tracking.
langchain AGENTS.md
Instructions for langchain-ai/langchain, covering global development guidelines for the langchain monorepo, corridor security analysis, project architecture and context, monorepo structure and development tools & commands.