feat: upgrade SAFE-UC-0025 (Enterprise agent-building platform) to draft#33
Merged
Conversation
…seed to full draft Expand seed into full draft covering the platform-of-platforms layer where enterprise customers build, deploy, govern, and operate their own AI agents. 7-stage kill chain with three NOVEL stages vs. sibling UCs (0018 / 0011 / 0021 / 0022 / 0024 / 0008): catalog-level tool poisoning, evaluation-harness bypass, and cross-tenant observability leakage. SAFE-MCP mapping across 24 techniques. Framework crosswalk spans NIST AI RMF + AI 600-1, EU AI Act Article 25 (provider/deployer routing), OWASP LLM Top 10 2025 (LLM03 supply chain + LLM07 system prompt leakage), ISO 42001/23894, NIST SP 800-218A, OpenSSF SLSA, and CSA MAESTRO. Incident citations: Tenable Copilot Studio SSRF (CVE-2024-38206), Noma ForcedLeak Salesforce Agentforce (CVSS 9.4), Aim EchoLeak M365 Copilot (CVE-2025-32711, distinguished as runtime not platform), Replit AIID #1152. Coined new workflow_family "Enterprise agent platforms & multi-tenant agent build". All citations live-verified in Phase 2 (80 URLs, 100% Tier A/B/C, zero Tier D). Signed-off-by: arjunastha <arjun@astha.ai>
Apply the new "Human technical-writer voice" skill rule that forbids em-dashes (U+2014) and double-hyphen-as-em-dash. Replacements were context-aware: markdown link titles use a colon, paired parentheticals use parens, and the common mid-sentence pattern becomes a period plus capitalization. Local validator passes; Phase 4.5 em-dash scan returns zero. No content changes beyond punctuation. Signed-off-by: arjunastha <arjun@astha.ai>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Promotes SAFE-UC-0025 from seed to full draft. Covers the platform-of-platforms layer where enterprise customers build, deploy, govern, and operate their own AI agents — distinct from runtime-agent UCs.
Changes
Citation accuracy
All 80 URLs independently live-verified in Phase 2 before commit (57 verification calls total). Tier coverage: 100% Tier A (regulators, standards bodies, government) / Tier B (vendor first-party docs, security-vendor disclosures, university research) / Tier C (reputable journalism with corroboration). Zero Tier D citations.
Precision-critical facts verified: SAFE-T1402 "Stenography" typo preserved verbatim; CVSS values correct (Tenable 8.5, ForcedLeak 9.4, EchoLeak 9.3); EU AI Act Article 25 effective date (2 August 2026, future); Replit AIID #1152 framed as operational failure not breach; M365 Copilot vs Copilot Studio distinction explicit; vendor rebrands annotated (Vertex AI Agent Builder → Gemini Enterprise Agent Platform; Mosaic AI Gateway → Unity AI Gateway; Claude Code SDK → Claude Agent SDK).
Safety attestation
No exploit steps, no sensitive information, defender-friendly throughout. Voice-drift scan returned 0 DRIFT hits — every `must` / `required` / `mandatory` falls into the hard-safety whitelist (tenant isolation, write-back gating, regulatory verbatim surfacing, attribution to human principal), factual-regulatory (quoting EU AI Act Article 25 verbatim), or structural template-inherited usage.
Requesting DSO review per CONTRIBUTING.md.