Subagent Verification for Claude AI Code Networks 2026
Automated Proof-of-Carrying Change Management for AIOps 2026
Practical and Efficient in-TEE Verification of Privacy Compliance
Deterministic proof-carrying decision core with replay verification, WAL, and fixed-point budget proofs
Every tool you've tried edits the document. That's why nothing holds — the document was never the source. Meaning is. Branch it, simulate a change before it exists, bind an approval to the exact state a human read, and let a merge preserve a conflict instead of inventing a number. Verified offline by a stranger. No model, no account. MIT.
Proof-carrying cyber immunity for Linux containers: signed evidence, deterministic policy, local approval and exact TTL containment.
Lean 4 formalization of Structural Explainability integration, composing substrate and identity regimes into cross-cutting theorems, traceability, and system-level invariants.
Logic Power v10 — proof-carrying decision logic: TLA+/Lean-verified gates, independent verifiers, reproducible artifacts
FLUX-C v3 VM — proof-carrying, SIMD-native, terminating constraint VM
Lean 4 formalization of identity regimes for Structural Explainability, defining six canonical regimes, nine derived profiles, and transformation-based identity classification.
Proof-carrying smart-contract auditing (AIS3 project): a finding must be an exploit the harness compiled and ran. Loses F1 to a tuned Bastet baseline; wins the pre-registered analyst-cost metric 8.7 vs 28.5 min per real bug.
Lean 4 formalization of the neutral structural substrate for Structural Explainability, defining admissibility, separation, and substrate invariants independent of identity regimes.
Proof-Carrying Operations: a dependency-free trust gate for agentic change. An AI agent edit to code, network configs, or Intune must carry a receipt proving identity, blast radius, and rollback, verified by a check it did not grade itself. Tuned for Claude Opus 4.8 and Codex.
Fail-closed Python runtime for AI agent authorization, seed-mediated authority migration, canonical JSON ledgers, and staged declared autonomy for RLHF-shaped agents.
Six single-focus Claude Code subagents for verifying changes to code, network, and Intune, built to Anthropic's subagent best practices, with a validator that proves the agent files conform. Drives the proof-carrying-ops suite. Stdlib only.
Observable-only workflow memory for long-running agents: promotes raw short-term traces into verified, receipt-bound workflow memory without relying on hidden meta-evaluators.
Toy framework illustrating limits of AI policy verification and the shift to proof-carrying, instance-level certification.
Proof-carrying CI gate for AI agent changes
To associate your repository with the proof-carrying topic, visit your repo's landing page and select "manage topics."