Memory Backfill

规范升级 Agent 记忆，将抽象原则转为包含项目事实和结果闭环的全面项目级记忆，并固化验收标准。

This skill appears internally consistent and focused on upgrading agent MEMORY files. Before installing, note: (1) it will read and write workspace memory files (MEMORY.md and memory/YYYY-MM-DD.md) — back up any sensitive records you don't want modified; (2) it asks for/uses real project names (with a note to desensitize), so ensure you follow your org's naming/privacy rules; (3) while the skill itself does not request network access or credentials, an agent executing these instructions might follow evidence paths (URLs or logs) if present — consider running in a test workspace or restricting the agent's network permissions if you are concerned; (4) review the referenced related skills (result-closure-memory, evidence-anchor, taskflow) before composing workflows that chain them. Overall the skill is coherent and low-risk, but validate in a controlled environment first.

Type: OpenClaw Skill Name: memory-backfill Version: 2.0.0 The memory-backfill skill is a structured methodology for an AI agent to manage and improve its internal memory files (e.g., MEMORY.md, SOUL.md). It provides templates and workflows for documenting project facts, decisions, and outcomes to ensure the agent's state remains accurate and evidence-based. The skill contains no executable code, external network requests, or instructions that deviate from its stated purpose of internal documentation management.