
from attentioncalculator111
Thin-client skill that queries a hosted service to generate rigorous integral proofs for 29 classes of mathematical inequalities (returns LaTeX proofs).
This skill is a thin client for a hosted attention-calculator service that generates integral-based proofs for a curated set of 29 inequality types (constants and functions like pi, e, trig/hyperbolic forms, gamma, Catalan, Apéry, etc.). It validates input, calls the remote /calculate and /get_integral_image endpoints, and returns the LaTeX integral proof for rendering. The skill enforces a strict IP/privacy boundary: no solver internals or coefficient tables are stored or reconstructed locally.
Use when a user requests a formal integral proof for an inequality belonging to the supported types. The skill should be invoked only after determining the request matches one of the supported proof types; otherwise it replies explaining the unsupported type (in Chinese). It's appropriate for mathematical assistants, proof-verification workflows, and educational tools that need machine-generated LaTeX proofs.
Best used by multilingual assistants that can make HTTP API calls and present LaTeX (e.g., Claude Code, agents with HTTP tools).
Thin-client skill that calls a hosted Chinese-language service (zhuyidao.net over plain HTTP) to generate integral proofs for 29 classes of mathematical inequalities, returning LaTeX. The bundled Python script is well-structured with proper argparse and error handling, but requires mandatory CLI arguments so DRY_RUN execution fails with a usage error (exit code 2). No security red flags beyond the use of plain HTTP instead of HTTPS for the remote API.
Legitimate thin-client math proof skill. No hidden payloads, no credential harvesting, no destructive commands. Main concern is plain HTTP for API calls, which is a privacy/integrity risk but not malicious. The skill explicitly instructs the agent never to reconstruct solver logic locally, which is an IP protection boundary rather than a security issue.