qimen039-code
UserModel-facing capability harness for LLM agent workflows: claim verification, memory continuity, risk routing, bounded improvement, and adapter contracts. DOI: 10.5281/zenodo.21189879.
Categories
Indexed Skills (5)
claim-boundary-harness
为豆包 Agent 提供 Claim Boundary Harness：声明边界、证据分级、风险路由、记忆隔离、事实/假设分离、输出复核和适配自检。用于严谨执行、长期记忆/跨对话接续、事实核查、发布前审查、R0-R5 风险判断、避免幻觉漂移或 CBH/claim-boundary 相关任务。
troubleshooting-skill-matrix
Router skill for a project-neutral agent self-improvement matrix. Use when routing semantic anchors, paired error records, paired solution records, external mechanism intake, or project router manifest rules.
agent-error-memory
Empty template for paired agent error memory. Use to record agent-caused execution, diagnosis, safety, routing, or process failures after they are solved and ready for reuse.
bug-solution-memory
Empty template for paired bug/problem solution memory. Use to record investigation order, fix path, rollback, validation, caveats, and reuse rules for solved incidents.
shared-semantic-anchors
Empty template for user-confirmed semantic anchors. Use only when a user explicitly defines or corrects a reusable meaning, boundary, naming rule, or interpretation shortcut.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.