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/aarwitz/tapp/agents-mdgit clone --depth 1 https://github.com/aarwitz/tappWhat 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 | $0.03044 | $0.03044 |
| Opus 5 | $0.01522 | $0.01522 |
| Sonnet 5 | $0.00609 | $0.00609 |
| Haiku 4.5 | $0.00304 | $0.00304 |
Grade A, and why
tapp AGENTS.md 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 — 172 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Tapp agent playbook
You (the agent) have Tapp: hands and eyes on real app surfaces — iOS simulators, Android emulators/devices, plus (beta) web apps in a real browser. Release judgment belongs only to Tapp's repository-connected deterministic gate.
No MCP connected? Just run the CLI
The core inspect, explore, replay, and gate capabilities work as plain commands — no Tapp account,
server, or global install. [target] is optional: with a repository/model, Tapp selects and prepares
one conclusive web, iOS, or Android target; it also accepts a repo dir, path/to/App.app, bundle id,
APK plus app id, or an owned HTTP(S) URL. You never need to know an iOS bundle id up front.
npx -y @aarwitz/tapp@latest explore [target] # autonomous exploration → findings + evidence (observation, not a gate; ≈ tapp_explore)
npx -y @aarwitz/tapp@latest focus "SCREEN OR CONTROL" [target] # source-locate + shortest observed route + screenshot
npx -y @aarwitz/tapp@latest explore https://your-app.example --watch # web: visibly follow the same exploration
npx -y @aarwitz/tapp@latest open [target] # launch + screen summary + screenshot saved to a file (≈ tapp_open_app)
npx -y @aarwitz/tapp@latest tree [target] # accessibility tree, --json for every element (≈ tapp_ui_tree)
npx -y @aarwitz/tapp@latest shot # screenshot the booted sim → file path (≈ tapp_screenshot)
npx -y @aarwitz/tapp@latest apps # what's installed on the simulator, with bundle ids
npx -y @aarwitz/tapp@latest build [dir] # build the app in an Xcode repo + install it (≈ tapp_build)
npx -y @aarwitz/tapp@latest explore app.apk --platform android --app-id com.acme.app
npx -y @aarwitz/tapp@latest flow run .tapp/flows/smoke.yml # committed, keyless E2E replay
If repository onboarding detects multiple application targets, target detection is deterministic but
the choice is the user's. Explicit init --explore asks even when the model has a saved default; a
later bare explore may consume that default. In a human TTY, Tapp displays a numbered selector and
continues in the same command. A non-interactive CLI prints the exact choices and exits before
building. MCP returns
reason: "target-selection-required" with structured choices[] (platform, name, sourcePath,
selector, and exact command). Do not pick one yourself. Present those choices to the user
with the client's native multiple-choice question UI when available, then rerun using the selected
--platform and --target. Plain chat can list the same choices when the client has no question
widget.
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 · 172 lines · 3,044 tokens per session scan A a929e4649e61
tapp AGENTS.md is an instructions file published in the GitHub repository aarwitz/tapp (0 stars, last pushed 3d ago), licensed MIT. It adds 3,044 tokens to every session, about $0.0152 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 2026-08-31.
Other instructions, from other repositories
vibe-testing CLAUDE.md
Claude Code instructions for AishwaryShrivastav/vibe-testing, covering vibe test — code-aware browser testing agent, what this project does, architecture, mcp tools (13) and development.
qaskills AGENTS.md
Instructions for PramodDutta/qaskills, covering agents.md, project overview, common commands, run a single package and run tests in watch mode.
agentic-playwright data-strategy.instructions.md
Instructions for idavidov13/agentic-playwright, covering data strategy, critical, file locations, instructions and phase 1: classify the value you need.
agentic-playwright config.instructions.md
Instructions for idavidov13/agentic-playwright, covering configuration, critical, file locations, instructions and phase 1: understand how env files are loaded.
inspecto CLAUDE.md
Instructions for inspecto-dev/inspecto, covering inspecto — claude code development guide, project overview, monorepo structure, development phases (load each file as needed) and key architectural decisions.
Stata-CLI AGENTS.md
Instructions for Utolaris/Stata-CLI, covering agents.md, project structure and working rules.