Universal Home Space Parser Engine （智能家居空间场景解析器）
v2.0.1S2 Official Smart Space Engine. Parses 62 spatial types into a 6-element hardware matrix. Includes a local MCP server and S2-SWM causality data harvester. /...
⭐ 0· 114·0 current·0 all-time
byMilesXiang@spacesq
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
medium confidencePurpose & Capability
The name/description (Universal Space Parser + SWM harvester) aligns with included Python parser, per-space dictionaries, an MCP server, a Chronos harvester, and a frontend. The code implements the advertised features (parsing 62 space types, local MCP tools, and writing causal logs). No unrelated services or credentials are requested.
Instruction Scope
The SKILL.md explicitly instructs running a local MCP server (s2_mcp_server.py) and writing causal event logs to s2_swm_training_data.jsonl. Those actions are within the declared purpose (world-model data harvesting + actuation), but they do grant runtime ability to receive MCP tool calls that simulate/perform physical actuation and persist S_t->A_t->S_t+1 records. If you run this on a machine that has physical device adapters, the MCP tools could be used to actuate them; the code currently simulates actuator calls but the design clearly expects adapter integration.
Install Mechanism
There is no opaque download or installer: the package is instruction/code-only. It requires Python 3.10+ and the 'mcp' Python package (the SKILL.md mentions pip install mcp). No URLs, extract steps, or remote binaries are embedded in the install spec.
Credentials
The skill requests no environment variables, no external credentials, and no config paths. The only runtime IO is local file writing (s2_swm_training_data.jsonl) and running an MCP server over stdio, which is proportional to its stated function of local causal data harvesting and LLM integration.
Persistence & Privilege
always:false and no modifications to other skills are declared. The skill persists data locally (the Chronos writer) but does not request permanent platform privileges. Autonomous invocation (model invocation) is allowed by default but that is the platform norm; combined with the MCP server capability this increases blast radius only if the host environment exposes real actuators.
Assessment
This skill appears internally consistent with its stated purpose, but it includes a local MCP server and a data harvester that will append S_t→A_t→S_t+1 entries to s2_swm_training_data.jsonl. Before installing or running: (1) review the s2_mcp_server.py and s2_chronos_memzero.py code to confirm there are no unintended network endpoints or adapter calls on your host; (2) run it in an isolated/test environment (no production devices attached) until you are confident; (3) if you do not want local logs, change the dataset filename/path or disable the Chronos write; (4) confirm the 'mcp' package source (pip package) and install it in a virtual environment; (5) consider file access controls or encryption for s2_swm_training_data.jsonl because it may contain sensitive sensor/state data; and (6) if you host a frontend/Next.js integration, avoid using untrusted server-side exec calls to run the Python script without sanitizing inputs. If you want a higher confidence assessment, provide the mcp package version/source or any adapters that would connect this server to real hardware so I can re-evaluate actuator risk.Like a lobster shell, security has layers — review code before you run it.
latestvk9709t1pa5qwas6wm149wgtgfh83xwp7
License
MIT-0
Free to use, modify, and redistribute. No attribution required.