Add RFSAM Skill: AI agent methodology for RF security audits (7-layer descent over 15 RF protocols)

[DragonJAR]

Summary

Adds a complete, self-contained Skill/ folder that turns any AI agent (Claude, ChatGPT, etc.) into a senior RF security auditor following the RFSAM methodology — a structured 7-layer descent (IG→SP→PHY→LL→CR→AT→AP) across 15 RF protocols, with an authorization gate, reproducible evidence workflow, and quality rubric.

Cross-fork PR from DragonJAR/RFSAM. The skill was developed and validated in the fork before opening this PR.

What Skill/ adds

Skill/
├── SKILL.md                  # Manifest + auth gate + 7-layer master flow
├── agents/openai.yaml        # Descriptor for OpenAI-compatible loaders
├── assets/
│   ├── finding-template.md        # Finding write-up template
│   ├── report-template.md         # Technical report template
│   └── executive-summary-template.md
├── references/
│   ├── 00-taxonomia.md            # Taxonomy, control IDs, coverage-map
│   ├── 01-autorizacion.md         # Legal frameworks by jurisdiction
│   ├── 02-kit-sdr.md              # SDR/sniffer hardware catalog
│   ├── 03-registro-hallazgos.md   # JSONL schema, CVSS 4.0 RF, severity model
│   ├── 10-ble.md … 24-uwb.md      # Per-protocol wayfinders (15 protocols)
│   ├── 25-troubleshooting.md      # Field diagnostics
│   └── 26-calidad.md              # Q1–Q8 quality rubric
└── scripts/
    ├── coverage_check.py          # Controls covered vs pending
    ├── register_finding.py        # Deterministic JSONL finding registration
    └── scaffold_report.py         # Report skeleton from JSONL

Why

RFSAM (Electronic Cats) is a web tool for RF security assessment methodology. This skill package makes the methodology portable to any AI assistant, so an auditor can load it and get structured guidance through the full 7-layer descent with protocol-specific commands, authorization gates, and reproducible evidence requirements — without needing the web app.

Authorization-first design

The skill enforces a hard authorization gate before any active step:

• 4 modes: Observational (RX only), Active (TX with re-check), Lab (TX + containment), Defensive (detect-only)
• 8 absolute rejections (RA1–RA8): surveillance of non-consenting third parties, jamming, spoofing critical infrastructure, rogue cells without license, credential cloning for fraud, etc.
• 5 redirects (RD1–RD5): web/API pentest, firmware dev, legal advice, forensics, hardware design

Not included

• No changes to the RFSAM web app code (only Skill/ folder + .gitignore)
• The skill guides and plans; real capture requires the operator's own hardware and authorization
• loot/ directory (captures, keys, findings) is gitignored and never committed

Validation

• All 3 Python scripts compile and pass self-tests (coverage_check.py, register_finding.py --self-test)
• Control IDs across all 15 wayfinders verified against canonical coverage-map
• Cross-references validated (no broken §N or references/XX-name.md links)
• Typo fixes applied: PBCH acronym correction (LTE + 5G NR), Zigbee crypto terminology correction, Wireshark spelling fix