Claude
|
d7d428d0a1
|
feat: Add gstack discipline, skill governance, Arabic ops layer
Final integration layer (gstack + Antigravity + Mukhtasar/Mkhlab):
- gstack_discipline.py: Planning enforcement with dispatch tiers
(Simple/Medium/Heavy/Full/Plan), plan validation, lite/full prompts
- skill_governance.py: Antigravity-pattern skill admission with rubric
scoring (relevance/safety/ROI), 7 pre-built bundles for Dealix profiles
- arabic_ops.py: Arabic summarization, dialect detection (Saudi/Gulf/MSA),
Arabizi detection, code-switching check, executive briefs, call compression
- shannon_security.py: Enhanced with verified findings and detailed PoC
- CLAUDE.md: Appended gstack tiers, Hermes profiles, Arabic ops guide
https://claude.ai/code/session_01LsnvBa7HwF5hs99VZbgLGj
|
2026-04-11 08:33:58 +00:00 |
|
Claude
|
1cebf54782
|
feat: Complete Hermes Fusion — execution router, Shannon, self-improvement, observability, API
Hermes Fusion Supreme integration:
- execution_router.py: Agent-level backend routing (Claude/OpenClaude/Goose/Internal)
with fallback chains, cost estimation, health tracking
- shannon_security.py: Staging-only white-box pentesting lane
(auth, injection, tenant isolation, PDPL compliance checks)
- self_improvement.py: Bounded inspect→measure→propose→verify→apply cycle
(max 5 proposals, max 2 auto-applies for trivial fixes)
- observability.py: Cost tracking, performance metrics, health monitoring,
Arabic executive summaries, anomaly detection
- hermes.py: Full API (execute, profiles, cost, health, improvements,
security scans, session restore — 18 endpoints)
https://claude.ai/code/session_01LsnvBa7HwF5hs99VZbgLGj
|
2026-04-11 08:29:09 +00:00 |
|