This comprehensive paper presents Model P3 (Compression Proxy), a methodology for passive classification of RSA public keys based on Algorithmic Compression Signatures (ACS). We validate that the compression ratio of a raw RSA modulus is not a stochastic variable, but a deterministic function of the key size and the fixed overhead of the compression algorithm (ZLIB, Zstd, or Brotli). Through rigorous experimental validation with N=150 samples generated in real-time, we demonstrate a Universal Scaling Law across multiple algorithms. The results fill the spectral gap with precision: RSA-2048 (R ≈ 1.043), RSA-3072 (R ≈ 1.029), and RSA-4096 (R ≈ 1.021). The variance was negligible (σ ≈ 0), establishing the metric as a reliable classifier for asset inventory. From an adversarial perspective, we demonstrate the tool's utility for Operational Reconnaissance Optimization rather than direct exploitation.