← Back to Skills Marketplace
tensorlake
by
Shanshan Wang
· GitHub ↗
· v2.6.1
· MIT-0
228
Downloads
0
Stars
0
Active Installs
9
Versions
Install in OpenClaw
/install tensorlake-skills
Description
Tensorlake SDK — guide for writing code that uses Tensorlake's sandbox product to build applications and AI agents. Use when the user mentions tensorlake or...
Usage Guidance
This skill is a documentation/authoring bundle for the Tensorlake SDK and is coherent with that purpose. Before acting on any of its recommendations: (1) prefer pip/npm installs over piping a remote installer into sh; if you consider `curl https://tensorlake.ai/install | sh`, review the script first and confirm the domain is trusted; (2) create and manage the TENSORLAKE_API_KEY in your environment or secret manager — do not paste keys into chat or commit them to source control; (3) when the agent shows SDK code, verify the symbols (imports/classes/methods) in your installed SDK or in the included references before executing; and (4) treat the CLI installer and any scripts it installs like third-party software — audit them if you need a higher assurance level.
Capability Analysis
Type: OpenClaw Skill
Name: tensorlake-skills
Version: 2.6.1
The bundle provides documentation and utility scripts for the Tensorlake SDK, a platform for sandboxed code execution and workflow orchestration. While the SDK enables high-privilege operations like interactive shells, desktop automation, and network egress control, these are aligned with the stated purpose of providing agent infrastructure. The bundle includes explicit security guardrails, such as instructions in SKILL.md and references/sandbox_advanced.md that forbid the agent from requesting API keys or modifying the user's host environment. No evidence of malicious intent, data exfiltration, or unauthorized persistence was found.
Capability Tags
Capability Assessment
Purpose & Capability
The name/description claim a Tensorlake SDK guide for sandboxes and orchestration, and the repository contents (SKILL.md, references/*, README, AGENTS.md) clearly implement that. There are no unrelated binaries, surprising env vars, or configuration paths requested by the skill itself.
Instruction Scope
SKILL.md scopes the agent to answering Tensorlake questions, producing SDK examples, and verifying symbols against bundled references. It also documents runtime prerequisites (TENSORLAKE_API_KEY) and shows SDK install commands. The doc suggests a CLI install using `curl -fsSL https://tensorlake.ai/install | sh` (a potentially risky pattern if blindly executed), but that is part of the SDK's published install instructions rather than hidden behavior in the skill. The instructions explicitly tell agents not to ask users to paste API keys into chat, which is good. Overall the instruction scope is appropriate, but the `curl | sh` pattern is worth flagging to the user as something to verify before running.
Install Mechanism
There is no skill install spec (instruction-only), which minimizes risk. The skill's documentation recommends installing the real Tensorlake SDK via pip/npm or running a curl|sh installer hosted on tensorlake.ai. The skill itself does not download or execute code during installation, but any user who follows the `curl | sh` recommendation should verify the installer source and prefer vetted package managers (pip/npm) or inspect the install script first.
Credentials
The skill does not declare required env vars in metadata, but the docs correctly state a single runtime credential: TENSORLAKE_API_KEY. That is expected and proportionate for an SDK that calls an external cloud API. No unrelated secrets or multiple credentials are requested by the skill.
Persistence & Privilege
The skill is not always-enabled, does not request persistent privileges, and contains only documentation plus benign helper scripts. It does not modify other skills or system-wide settings.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install tensorlake-skills - After installation, invoke the skill by name or use
/tensorlake-skills - Provide required inputs per the skill's parameter spec and get structured output
Version History
v2.6.1
- Updated SDK version to 2.6.1.
- Improved wording and streamlined documentation for clarity and conciseness.
- Added default resource parameters (image, CPU, memory, disk, timeout) to example sandbox creation.
- Updated snapshot terminology to "filesystem (default)" and "memory" snapshots and clarified snapshot usage.
- Expanded framework integration examples to include Claude and OpenAI agent SDKs.
- Minor formatting and organizational improvements for better readability.
v2.6.0
- CLI support added: now includes CLI installation instructions and usage alongside Python and TypeScript.
- Quick Start updated: initial example now demonstrates launching a sandbox and interacting with files, replacing workflow orchestration example.
- Expanded core sandbox patterns: covers parallel snapshot forking, interactive PTY shells, port exposure, and custom images.
- Orchestration guidance split out and restructured; separate Python, TypeScript, and CLI notes.
- Miscellaneous improvements: clarified API references, and updated documentation links and setup steps.
v2.5.5
tensorlake-skills 2.5.5
- Expanded and clarified the SKILL description to explicitly cover additional sandbox features (interactive shells, PTY, port exposure, egress allowlists, desktop automation, file transfer).
- Added documentation of "computer use / desktop automation" pattern using VNC sandboxes and keyboard/mouse control.
- Updated pattern descriptions to improve guidance on when to use sandboxes for agentic applications and code execution.
- No code or structural file changes; changelog is for documentation and instructions only.
v2.5.4
- Bumped version to 2.5.4.
- Added CLAUDE.md reference file.
- Updated core pattern notes: clarified snapshot restore behavior by distinguishing between "filesystem" (default) and "full" snapshot types and explaining resource override rules (CPU, memory, disk).
- Instructed to check `Sandbox.get_snapshot(snapshot_id).snapshot_type` or the dashboard before advising on rebuilding from scratch to change resources.
v2.5.1
- Updated documentation to reference the new CLI entrypoint name (`tl`) and clarified aliasing for `tensorlake`.
- Expanded "Core Patterns" to explicitly cover LLM code-execution workflows in sandboxes, with new best practices and note on process state persistence.
- Improved CLI usage instructions throughout, reflecting preferred CLI usage (`tl`) and associated commands.
- Removed the CLAUDE.md reference file.
v2.4.1
- Added guidance to verify all Tensorlake SDK symbols and import paths before using or suggesting them.
- Instructs to decline symbol suggestions if their existence cannot be confirmed.
- No functional changes to the SDK usage or API described.
- Minor documentation clarification—core setup and usage instructions remain unchanged.
v2.4.0
- Updated Orchestrate workflow/API naming to "Orchestration" for consistency.
- Minor edits and clarifications to descriptions and quick start documentation.
- No changes to CLI commands or SDK features.
- Version bump from 2.3.0 to 2.4.0.
v2.3.0
- Updated environment variable instructions to clarify that the only required variable is TENSORLAKE_API_KEY, and to always use this exact name.
- Improved documentation for secret management and environment variable configuration, specifying provider keys are optional and user-scoped.
- Added LICENSE file to the repository.
- Bumped version to 2.3.0.
v2.2.0
- Expanded, detailed documentation for Tensorlake SDK usage, setup, and orchestration patterns.
- Added clear instructions for configuring API keys and using CLI tools securely.
- Introduced comprehensive Quick Start and Core Patterns sections for both Python and TypeScript.
- Provided rules and code examples for composing agentic workflows with sandboxed execution.
- Bundled specific reference files for Orchestrate and Sandbox APIs, document extraction, integrations, advanced usage, and troubleshooting.
- Linked to the official Tensorlake docs for further detail beyond packaged references.
Metadata
Frequently Asked Questions
What is tensorlake?
Tensorlake SDK — guide for writing code that uses Tensorlake's sandbox product to build applications and AI agents. Use when the user mentions tensorlake or... It is an AI Agent Skill for Claude Code / OpenClaw, with 228 downloads so far.
How do I install tensorlake?
Run "/install tensorlake-skills" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is tensorlake free?
Yes, tensorlake is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does tensorlake support?
tensorlake is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created tensorlake?
It is built and maintained by Shanshan Wang (@cooleel); the current version is v2.6.1.
More Skills