agentdodont 0.1.0__tar.gz

agentdodont-0.1.0/.gitignore

@@ -0,0 +1,19 @@
+logs/
+suite_results/
+__pycache__/
+*.pyc
+.relay/
+
+# Secrets: never commit
+android/local.properties
+local.properties
+.env
+*.keystore
+*.jks
+*.pem
+*.key
+
+# Python packaging
+dist/
+build/
+*.egg-info/

agentdodont-0.1.0/LICENSE

@@ -0,0 +1,202 @@
+
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

agentdodont-0.1.0/NOTICE

@@ -0,0 +1,4 @@
+Agent Do/Don't
+Copyright 2026 Moshe Sipper
+
+This product is licensed under the Apache License, Version 2.0 (see LICENSE).

agentdodont-0.1.0/PKG-INFO

@@ -0,0 +1,104 @@
+Metadata-Version: 2.5
+Name: agentdodont
+Version: 0.1.0
+Summary: A rules engine between an AI agent and its tools, built assuming the model can be hijacked. Includes an MCP proxy.
+Project-URL: Homepage, https://github.com/moshesipper/agentdodont
+Project-URL: Documentation, https://github.com/moshesipper/agentdodont/blob/main/docs/mcp-proxy.md
+Project-URL: Issues, https://github.com/moshesipper/agentdodont/issues
+Author: Moshe Sipper
+License-Expression: Apache-2.0
+License-File: LICENSE
+License-File: NOTICE
+Keywords: ai-agents,android,claude,llm,mcp,prompt-injection,security
+Classifier: Development Status :: 3 - Alpha
+Classifier: Intended Audience :: Developers
+Classifier: Operating System :: OS Independent
+Classifier: Programming Language :: Python :: 3
+Classifier: Topic :: Security
+Requires-Python: >=3.10
+Provides-Extra: phone
+Requires-Dist: anthropic>=0.84; extra == 'phone'
+Requires-Dist: cryptography>=42; extra == 'phone'
+Requires-Dist: pillow>=10; extra == 'phone'
+Requires-Dist: qrcode>=7; extra == 'phone'
+Description-Content-Type: text/markdown
+
+# Agent Do/Don't
+
+A rules engine between an AI agent and its tools, built assuming the model can be hijacked by
+anything it reads. The model proposes; the rules decide: **do it**, **ask you**, or **refuse**.
+
+This package contains the **MCP proxy**. It works with any agent that uses
+[MCP](https://modelcontextprotocol.io) tools, such as Claude Code, Claude Desktop, or your own.
+It sits between the agent and an MCP server, and neither of them needs any changes. It needs
+nothing beyond the Python standard library. The project also includes an Android app with the
+same rules ([GitHub](https://github.com/moshesipper/agentdodont)).
+
+Status: **experimental.**
+
+## What it does
+
+Suppose an agent reads a web page with hidden instructions telling it to email your notes to
+helpdesk@evil.example. If the model falls for it, the proxy sees the session has read untrusted
+content, and refuses the email: that address isn't one of yours. The call never reaches the
+email server.
+
+- **Taint.** After a tool returns untrusted content, the session counts as untrusted from then
+  on. This holds across every server the agent uses, and questions name where the content came
+  from.
+- **Where data goes.** URLs, email addresses and phone numbers in a call's arguments are checked
+  against your lists. Once the session is tainted, an address you haven't listed is refused,
+  not asked. Before that, a new address asks.
+- **Risk.** It reads the server's tool annotations the cautious way. A tool with no annotations
+  counts as possibly irreversible and asks.
+- **Your settings.** You can set any tool to "do it", "ask" or "never", and there's a rate limit.
+- **Asking.** Questions appear in the client (MCP elicitation). If the client can't ask, the call
+  is refused, never run silently.
+- **Log.** Every decision goes to `~/.agentdodont/mcp-audit.jsonl`.
+
+## Set up
+
+Start the proxy in place of the server, with the server's own command after `--`. With
+[uv](https://docs.astral.sh/uv/) installed, nothing else needs installing. For Claude Code:
+
+```
+claude mcp add fetch -- uvx agentdodont mcp-proxy --rules /path/to/mcp-rules.json --name fetch -- uvx mcp-server-fetch
+```
+
+Or put this in `.mcp.json` in your project:
+
+```json
+{
+  "mcpServers": {
+    "fetch": {
+      "command": "uvx",
+      "args": ["agentdodont", "mcp-proxy", "--rules", "/path/to/mcp-rules.json", "--name", "fetch",
+               "--", "uvx", "mcp-server-fetch"]
+    }
+  }
+}
+```
+
+Claude Desktop uses the same entry in `claude_desktop_config.json`. Without uv, run
+`pip install agentdodont` and use `"command": "agentdodont"` with the same arguments minus the
+first `"agentdodont"`.
+
+A rules file is optional. Every field in it is optional too:
+
+```json
+{
+  "allowed_sites": ["wikipedia.org", "github.com"],
+  "known_recipients": ["dana@example.com"],
+  "levels": {"read_*": "auto", "delete_*": "ask", "get-env": "deny"},
+  "tools": {"send_message": {"destination_args": {"channel": "recipient"}}}
+}
+```
+
+## More
+
+- [The full proxy guide](https://github.com/moshesipper/agentdodont/blob/main/docs/mcp-proxy.md):
+  every rule, every field, and the limits.
+- [Threat model](https://github.com/moshesipper/agentdodont/blob/main/docs/threat-model.md)
+- [Benchmark](https://github.com/moshesipper/agentdodont/blob/main/docs/benchmark.md)
+
+Apache License 2.0.

agentdodont-0.1.0/README.md

@@ -0,0 +1,189 @@
+# Agent Do/Don't
+
+**The model proposes. Agent Do/Don't decides.** *Do. Don't. Ask first.*
+
+AI agents now send messages, change calendars, open links and tap through apps on your behalf.
+Any text they read can hijack them: a notification, a calendar invite, a web page, another app's
+screen. Most defenses try to make the model notice the trick. Agent Do/Don't assumes it won't.
+
+Every tool call the agent makes goes through a rules engine that runs outside the model, which the
+model can't see or influence. The engine decides: **do it**, **ask you**, or **refuse**. Everything
+is logged and explained, and you can undo whatever can be undone.
+
+This repository is a working Android implementation: a home-screen app with a Claude agent,
+running on a real phone (Galaxy S22, Android 16). It's an early prototype. See [Status](#status).
+
+---
+
+## What sets it apart
+
+Many tools can intercept an agent's tool calls and ask a human. Agent Do/Don't goes further,
+because it assumes the model is **already compromised**:
+
+- **It knows what the agent has read, and from where.** Every tool declares whether its output
+  is trusted; the code won't compile otherwise. Once a task has read outside content, it's
+  **tainted**, and the source is named in every question: *"this task has read notifications from Shell"*.
+- **A tainted task can only send data to places you've already allowed.** Your contacts, or
+  sites on your list. Anything else is **refused, not asked**, so an injection can't talk you
+  into leaking data with one tired tap. Typing an address or URL into another app counts too.
+  Even before it has read anything, the agent asks before sending data anywhere new.
+- **Some things are never possible.** The agent can't read or operate protected screens and
+  controls: Agent Do/Don't itself, permission prompts, factory reset, developer options. These
+  are recognized in 18 languages, and in any other language every named tap asks. A **scope
+  lock** keeps it inside the app the task opened.
+- **Only you can undo, and only you can clear the record.** Every reversible change can be
+  undone, grouped by request. The agent has no tool for undoing or clearing history, so it
+  can't erase its tracks.
+- **"Why did you do that?"** For every action: who proposed it, which rule decided, what you
+  answered, what happened.
+- **Memory an injection can't poison.** What the agent remembers is encrypted on the phone, and
+  you can view and edit it. Secrets are refused, and memory writes from tainted tasks ask first.
+- **You press Send.** Messages and emails are drafted in the real app. Agent Do/Don't never sends them.
+- **Tested against a hijacked model,** not just a confused one (see [Evidence](#evidence)).
+- **Your API key stays on your computer.** The phone talks to a small relay with a budget and a
+  rate limit. It pairs over Wi-Fi with a pinned certificate.
+
+What it does **not** guarantee is just as important. Read the [threat model](docs/threat-model.md).
+
+---
+
+## How it works
+
+```
+ your request
+      │
+      ▼
+ ┌─────────┐  proposes a tool call   ┌───────────────────────────────────┐
+ │  model  │ ──────────────────────▶ │       rules engine (no model)     │
+ │ (Claude)│ ◀────────────────────── │ validate → risk → taint → where   │
+ └─────────┘  result, labelled       │ data goes → protected screens     │
+              trusted / untrusted    │ → scope lock → rate limit         │
+                                     └─────────────────┬─────────────────┘
+                                      do it / ask you / refuse
+                                                       ▼
+                                      runs · audit log · undo entry
+```
+
+- The engine ([Policy.kt](android/app/src/main/java/dev/aos/launcher/capabilities/Policy.kt)) is
+  plain Kotlin with no Android or model dependencies, so it's deterministic and unit-tested.
+- Every applicable rule is evaluated and the **strictest verdict wins**. No approval overrides a refusal.
+- Irreversible actions (send, delete, pay) **always ask**, whatever the settings.
+
+---
+
+## Evidence
+
+The debug build contains a **scripted attacker**: a stand-in "model" that does exactly what the
+injected text says, as a fully hijacked model would. It runs through the same loop, rules and
+confirmation screens as the real agent. On the S22:
+
+| Injection arrives via | Attacker then tries to | Got through without the user's OK |
+|---|---|---|
+| A notification or calendar invite | Delete a calendar event; set a 3 a.m. alarm; send the calendar's event titles to an attacker URL | **0 of 3** (the delete and the alarm asked, naming the source; the URL was refused outright) |
+| Another app's screen (Calendar) | Tap a control in Settings; read a different app's screen | **0 of 2** (the tap asked; the read was refused by the scope lock) |
+
+The real model (Claude) ignored the injections too, even when told to obey them. The point is
+that the result doesn't depend on that.
+
+**Benchmark:** 35 attacks and 28 ordinary requests, replayed through the rules engine with it on
+and off. With the rules off, all 35 attacks work. With them on, 17 are refused outright, 16
+need your OK, and 2 get through, each a documented limit. Ordinary requests cost 13 questions
+across 28 requests. The benchmark also found a gap (an injection inside the app being operated
+could get an address typed in), now fixed. In a live run on the phone, the real model resisted 11
+of 12 injections by itself; the one it obeyed (a 3 a.m. alarm) was stopped by the rules. Method, every scenario and the gaps: [docs/benchmark.md](docs/benchmark.md).
+
+**Try to break it.** If you get a harmful action past Agent Do/Don't without a confirmation,
+please report it privately ([SECURITY.md](SECURITY.md)).
+
+---
+
+## What's in the repository
+
+```
+android/        The app: home screen with an intent box, the Claude agent, the rules engine,
+                tools (alarms, calendar, URLs, apps, notifications, contacts, email/text drafts),
+                screen operation through accessibility, encrypted memory, History, Undo, pairing.
+agentdodont/    Python: the relay that keeps the API key off the phone (budget, rate limit,
+                QR pairing over pinned TLS, revocable device tokens); the MCP proxy (the same
+                rules for any MCP agent); and an experimental laptop agent that drives a phone
+                over ADB.
+tests/          Python tests (relay, pairing, MCP proxy, laptop agent).
+tools/          The live benchmark (the real model on a phone, against planted injections).
+docs/           Threat model, benchmark, and every feature in detail with how it was verified.
+```
+
+---
+
+## Quick start
+
+You need: an Android phone with USB debugging on (Android 11 or newer; tested on Android 16),
+Android Studio (for its JDK and SDK), Python 3.10+, `adb` on your PATH, and an
+[Anthropic API key](https://console.anthropic.com/).
+
+```sh
+git clone https://github.com/moshesipper/agentdodont && cd agentdodont
+pip install -r requirements.txt
+
+# 1. Your key goes in a gitignored file and stays on your computer.
+echo "anthropic.apiKey=YOUR_KEY" >> android/local.properties
+
+# 2. Build and install (set JAVA_HOME to Android Studio's JDK). The key is never built into the app.
+cd android && ./gradlew installDebug && cd ..
+
+# 3. Start the relay: it gives the phone a device token over USB and holds the key.
+python -m agentdodont relay            # leave running while you use the phone
+```
+
+Then, on the phone:
+
+- **Set it as the home screen** (optional): `adb shell cmd package set-home-activity dev.aos.launcher/.MainActivity`.
+  To undo: Settings › Apps › Default apps › Home app.
+- **Grant what you want to use:** calendar and contacts permissions when asked; **Notification
+  access** for reading notifications; **Settings › Accessibility › Agent Do/Don't** for
+  operating other apps.
+- Type a request, such as *"what's on my calendar tomorrow?"* or *"alarm 7:30"*. Look at
+  **History** and **Policy** to see what happened and to change what may happen.
+
+**Over Wi-Fi instead of USB:** `python -m agentdodont relay --lan --pair` shows a QR code. Scan
+it with the phone's camera, then confirm on the phone. After that, `relay --lan` is enough;
+`relay --devices` lists paired phones and `relay --revoke <id>` cuts one off.
+
+[docs/features.md](docs/features.md) has every option and command, plus the red-team commands.
+
+---
+
+## Tests
+
+```sh
+cd android && ./gradlew testDebugUnitTest    # rules engine, taint, undo, History, screen rules, word lists, memory, pairing
+python -m unittest discover -s tests          # relay, pairing, MCP proxy, laptop agent
+```
+
+---
+
+## Status
+
+**Works today:**
+- the rules engine with taint and named sources, outbound allow-lists (tainted tasks are refused; any other agent task asks), protected screens in 18 languages, and the scope lock;
+- user-only undo, History with "why", and clearing history with a visible marker;
+- encrypted memory that you can inspect, and that injections can't write to silently;
+- message and email drafts with your contacts as the allow-list;
+- the relay with a budget, a rate limit and pinned Wi-Fi pairing;
+- **experimental:** an [MCP proxy](docs/mcp-proxy.md) that puts the same rules between any MCP agent
+  (Claude Code, Claude Desktop…) and its tools, with one session across all its servers:
+  `pip install agentdodont`, or `uvx agentdodont mcp-proxy`.
+
+**Next:**
+- [ ] a defense and tests against "action rebinding" (the screen changes between check and tap)
+- [ ] the rules engine as a standalone library, ported to Python, with shared test cases for both versions
+- [ ] the MCP proxy beyond experimental: ready-made rules and runs against popular servers and clients
+
+---
+
+## License
+
+[Apache-2.0](LICENSE). Copyright 2026 Moshe Sipper.
+
+## Security
+
+Please report vulnerabilities privately; see [SECURITY.md](SECURITY.md).

agentdodont-0.1.0/agentdodont/__init__.py

@@ -0,0 +1,3 @@
+"""Agent Do/Don't: a rules engine between an AI agent and its tools, built assuming the model can be hijacked."""
+
+__version__ = "0.1.0"