Evidence-driven reverse engineering and unpacking workflows for AI agents, with reproducible Windows experiments.
Make AI coding agents architecture-aware: baseline-first, evidence-verified, drift-checked, and safe across long tasks.
A Seedance 2.5 TVC director Skill for 15s, 30s and custom commercial video prompts, storyboards, voiceover routing and evidence-driven retakes.
面向 AI/CS 科研的多 Skill 实验智能体:助力全自动科研实验,以 Claim 驱动、骨干实验优先、独立结果审计与 Evidence Freeze,把发散的模型能力收束为可追溯、可复现、可收敛的论文证据。
Evidence-driven autonomous SOC team on AgentTeams: hash-bound approvals, HMAC audit chain, rollback and reproducible attack scenarios — auditable governance primitives for agent infrastructure.
Evidence-driven, privacy-first reflective automation for personal agents.
Evidence-driven production readiness checklist for shipping web applications with confidence
Evidence-driven Product Manager Workflow Skill for Codex
Evidence-driven forensic investigation: how a near-perfect A-share ML prediction signal collapsed under rigorous validation
Evidence-driven SGLang inference deployment optimization.
Evidence-driven Java security review workbench
Agent-native embedded systems lab for pluggable simulation, evidence-driven validation, and controlled evolution.
Evidence-driven Codex skill for reverse-engineering product journeys, panorama architecture, and Agent contracts.
Bounded, evidence-driven SKILL.md evolution under frozen evaluation and safety contracts.
Evidence-driven HiMCM modeling skills for Codex & Claude Code: reproducible experiments, verified figures, frozen results, and audited papers.