Smalltalk

Interact with live Smalltalk image (Cuis or Squeak). Use for evaluating Smalltalk code, browsing classes, viewing method source, defining classes/methods, querying hierarchy and categories.

What to consider before installing: - Source review: the code is self-contained and matches the described purpose, but SKILL.md asks you to git-clone an external GitHub repo; review that repo yourself before running to ensure you trust it. - Local VM and files: this skill starts a local Squeak VM (via xvfb-run), creates background daemons, and reads/writes image and .changes files — do not run it with images you don't trust, and be prepared for large image files to be copied into ~/smalltalk-projects. - Remote services: explain/audit features optionally use Anthropic/OpenAI keys; only set those env vars if you accept sending method/source text to those providers. - Persistence & cleanup: the daemon leaves sockets/pid files in /tmp and stores projects under ~/.clawdbot and ~/smalltalk-projects; stop the daemon and remove those files if you want to revoke persistence. - Dangerous actions are explicit: commands can define/delete classes and methods in the live image. That functionality is consistent with the tool but is powerful — avoid running untrusted code or images. If you want higher assurance, inspect the full Python sources included here, avoid cloning or running untrusted images, and only provide LLM API keys when needed.

Type: OpenClaw Skill Name: smalltalk Version: 1.7.0 This skill is classified as suspicious due to its high-risk capabilities, including arbitrary Smalltalk code execution (via `evaluate`, `define-class`, `define-method`, etc.) and the execution of LLM-generated Smalltalk code (in `smalltalk.py` via `tool_generate_sunit`). While these features are central to the skill's stated purpose of interacting with a live Smalltalk image, they present a significant attack surface for code injection and potential VM escape. Additionally, the `debug_squeak` function in `smalltalk.py` creates an HTML report containing screenshots and stack traces in `/tmp`, posing a local information disclosure vulnerability. The `smalltalk-daemon.py` manages a persistent VM with powerful capabilities, though it implements some security measures like user-isolated Unix sockets with restrictive permissions. The skill does not show clear evidence of intentional malicious behavior like data exfiltration to external endpoints or stealthy persistence mechanisms.