feat: discover-environment skill + diagram fixes + agent-bom CI

[msaad00]

Summary

New skill: discover-environment

• Maps cloud resources to security graph with MITRE ATT&CK (9 techniques) and ATLAS (6 techniques) overlays
• Supports: AWS (boto3), GCP (google-cloud), Azure (azure-mgmt), static config (JSON/YAML)
• Output: standalone graph JSON — no agent-bom dependency, any tool can consume it
• Entity types: user, service_account, credential, cloud_resource, server (Lambda/Functions)
• Relationships: contains, owns, uses, exploitable_via
• 15 tests covering graph model, MITRE mapping, static config, stats

Diagram fixes

• Simplified model-serving and GPU cluster diagrams — removed spaghetti edges
• Clean linear flow: Config → Benchmark → Output (no every-to-every connections)
• All diagrams use consistent slate/teal/cyan/indigo palette (no red, no green)

CI additions

• test-discover-environment job
• agent-bom-scan — pip install + CLI for reliability (replaced GitHub Action)
• agent-bom-skills-audit — skill definition security review
• Both agent-bom jobs: continue-on-error: true, results uploaded as artifacts

Test plan

• discover-environment: 15 tests passing
• ruff check + format: all clean
• Diagrams simplified — no spaghetti edges