Results for “8k4-protocol”

How MaksPM hands off work to agents using sessions_spawn for continuous pipeline flow, with file-reference pattern for preserving full project specs. Updated 2026-03-21 with 9-phase pipeline and Forge architecture phase.

Workflow 4: Submission rebuttal pipeline. Parses external reviews, enforces coverage and grounding, drafts a safe text-only rebuttal under venue limits, and manages follow-up rounds. Use when user says "rebuttal", "reply to reviewers", "ICML rebuttal", "OpenReview response", or wants to answer external reviews safely.

Guides a structured planning pipeline for coding tasks, from scoping and research to plan creation, review, and handoff.

Provides production-grade Kubernetes YAML patterns and kubectl debugging commands for deploying, managing, and debugging workloads reliably.

Generates an operations runbook from deployment configs, CI/CD pipelines, and monitoring setup, with copy-pasteable procedures for deployment, rollback, scaling, database maintenance, incident response, and troubleshooting.

Integrate Hardware Security Modules (HSMs) using the PKCS#11 interface for cryptographic key management, signing operations, and secure key storage with python-pkcs11, AWS CloudHSM, and YubiHSM2.

Guides a structured planning pipeline for coding tasks, including scope detection, research, plan creation, and review.

Runs a structured bug-fix pipeline: scouts evidence, diagnoses root cause, applies minimal fixes, and reviews before committing.

Delegates a bounded repository task to an allowlisted external coding CLI through a guarded runner, with provider selection, timeout handling, and independent local review.

Meta's 7-8B specialized moderation model for LLM input/output filtering. 6 safety categories - violence/hate, sexual content, weapons, substances, self-harm, criminal planning. 94-95% accuracy. Deploy with vLLM, HuggingFace, Sagemaker. Integrates with NeMo Guardrails.

Humanization Pipeline Orchestrator v3.1 - Multi-pass 4-layer transformation pipeline Orchestrates G5 (Auditor), G6 (Humanizer), F5 (Verifier) in sequential passes Enforces checkpoints between every pass with mandatory AskUserQuestion Supports conservative (L1-2), balanced (L1-3), balanced-fast (L1-3 merged), aggressive (L1-4) modes Rich Checkpoint v2.0: section-level scores, selective humanization, target auto-stop G5+F5 parallel execution, section-selective humanization Triggers: humanize, humanize my draft, humanize manuscript, make natural, remove AI patterns Korean triggers: 휴먼화, 자연스럽게, AI 패턴 제거

Meta's 7-8B specialized moderation model for LLM input/output filtering. 6 safety categories - violence/hate, sexual content, weapons, substances, self-harm, criminal planning. 94-95% accuracy. Deploy with vLLM, HuggingFace, Sagemaker. Integrates with NeMo Guardrails.

Implements a feature phase by phase from a phased JSON or Markdown plan, with configurable unit-test and quality-gate choices per phase, token-budgeted execution, resumable state, and TDD handoff.

Diagnoses and operates Kubernetes clusters: pods, deployments, ingress, HPA, PVC, StorageClass, NetworkPolicy, Secrets, Helm, ArgoCD, cert-manager. Reads before changing; mutations require confirmation and diff review.

AWS CDK infrastructure builder using TypeScript with L2/L3 constructs and Well-Architected patterns. Activate on: AWS CDK, CDK construct, CDK stack, CDK pipeline, AWS infrastructure as code TypeScript, L2 construct, CDK patterns. NOT for: Terraform IaC (use terraform-module-builder), Kubernetes manifests (use kubernetes-manifest-generator), serverless framework (use devops-automator).