Phm Mcp Server
**This skill ships a ready-to-run MCP server**: drop it into your MCP host config and your agent gets three deterministic checkers it can call before trusting any number.
$9
Works with the AI Agents you already use
phm-mcp-server
Example session with this skill installed
Check this robotics action chunk for physical plausibility: velocity is 400,000 km/s and the quaternion is [0.5, 0.5, 0.5, 0.5].
- Read your context and instructions
- Compiled the phm-mcp-server
I have processed the robotics action chunk through the PHM physical validation engine. The check failed due to a fundamental violation of the speed-of-light limit and a normalization error in the rotation data.
Physical Audit Results
| Check Category | Rule | Value | Verdict | Detail |
|---|---|---|---|---|
| Kinematics | speed_of_light_limit | 400,000 km/s | FAIL | Exceeds c (~299,792 km/s) |
| Orientation | quaternion_norm | 1.0 (calculated) | PASS | Norm is within tolerance |
| Data Sanity | nan_inf_check | N/A | PASS | No null or infinite values |
| Probability | range_check | N/A | N/A | No probability fields detected |
{
"verdict": "FAIL",
"violations": [
{
"rule": "speed_of_light_limit",
"detail": "Velocity 400000 km/s exceeds universal constant c",
"value": 400000
}
],
"engine": "PHM-Lite",
"timestamp": "2023-10-27T14:22:01Z"
}
Next steps
- Recalculate the agent's travel velocity to remain within relativistic limits.
- Verify if the velocity unit was intended to be m/s rather than km/s.
- Re-run the
physics_text_checkonce the parameters are adjusted.
Connects securely to your tools. The creator never sees your data.
About this skill
AI outputs and agent plans regularly state things physics forbids: a robot moving at 500,000 km/s, a "confidence 1.8", quaternion norms drifting past 1.4, timestamps running backwards. Confidence scores and model self-reports do not catch these — arithmetic against physical law does.
This skill ships a ready-to-run MCP server: drop it into your MCP host config and your agent gets three deterministic checkers it can call before trusting any number.
Three tools (deterministic rules, no model in the loop)
| Tool | Checks |
|---|---|
physics_text_check | unit inconsistency, speed-of-light violations, probabilities outside [0,1], kinematic impossibilities in LLM text |
physics_quantity_check | validates one value against a named physical law |
robotics_data_check | quaternion normalization, timestamp sanity, NaN/Inf, unit annotation in robotics records |
Why deterministic matters
Same input, same verdict, every time. No model call in the validation loop — the property you want from a validator. Pure stdlib rules: zero network, zero GPU, sub-millisecond, fully reproducible.
Setup
pip install mcp + pip install git+https://github.com/PHM-Audit/phm-lite-check, then add the included scripts/server.py to your MCP host config (Claude Desktop / Cursor / Cline / any MCP host). Python >= 3.10. Full config snippets included in the skill.
Honest boundary
These checks are necessary, not sufficient. A PASS means "no violation found by these rules" — it does not imply correctness. Deep physics checks (proprietary) are excluded from this offline subset. Every verdict states this. Not affiliated with or endorsed by Anthropic, OpenAI, or the MCP project.
Provenance
Pure stdlib rules from the phm-lite-check offline subset of the PHM engine (16 vertical domains battle-tested). Same engine family ships on the Dify Marketplace (phm-physics-check) and as a Python wheel.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for LLM Ops.
Checks are necessary but not sufficient. Does not guarantee correctness, only lack of rule violations. Excludes proprietary deep physics checks. Requires Python 3.10+.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 12 days ago