Document consolidated tested GCP attack techniques

[carlospolop]

Summary

This consolidates and supersedes #378, #380, #381, and #382 in the current master / src documentation layout.

Privilege escalation and trust restoration

• Workload Identity Federation SAML provider creation and provider takeover
• deleted-provider restoration
• disabled/deleted pool restoration
• exact-permission REST paths and no-list fallbacks

GCP ↔ Workspace and credential pivots

• Workspace HTTP add-on deployment endpoint takeover
• Agent Identity credential retrieval and OAuth token-endpoint interception
• Application Integration decrypted auth-config access and invoke-only credential forwarding
• cross-project Connector runtime-identity attachment prerequisites and tested boundaries
• Workspace group-to-GCP membership boundary clarification

Runtime and source/configuration disclosure

• project, zonal, VM, instance-template, and machine-image Compute metadata
• App Engine literal environment variables, runtime identity, and deployment locations
• Vertex AI complete pipeline specifications, parameters, commands, arguments, and execution identity

Validation

• all included attack paths retain their original isolated live-test evidence and cleanup boundaries
• twelve runtime-disclosure permissions were independently isolated with matching success and opposite GET/LIST HTTP 403 controls
• local links, SUMMARY targets, numbered references, HTML/details blocks, hint pairs, and fenced blocks validated
• git diff --check passed
• full mdbook build passed with the repository-pinned translator-image; generated book/ output remains excluded
• branch is rebased on the latest master

Supersedes #378, #380, #381, and #382.