菜单
Sage's persistent memory. Use when the user says remember this, store this, don't forget, note for next time, what do we know about X, have we seen this before — and whenever the user corrects you, an approach fails repeatedly, or you discover a non-obvious gotcha worth keeping.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
基于 SOC 职业分类
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/xoai/sage --skill sage-using-memory命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
正在显示 SKILL.md
| name | sage-using-memory |
| description | Sage's persistent memory. Use when the user says remember this, store this, don't forget, note for next time, what do we know about X, have we seen this before — and whenever the user corrects you, an approach fails repeatedly, or you discover a non-obvious gotcha worth keeping. |
| version | 1.0.0 |
| type | system |
Sage has its own memory. It is not the platform's native memory
(MEMORY.md, feedback files, and the like). Storing a Sage learning in the
platform's memory puts it where no other Sage session will look for it.
Storage target: the sage_memory_store() MCP tool.
Fallback: .sage-memory/ markdown files when the MCP server is unavailable.
Before writing a spec, a plan, or an ADR — or before starting an investigation — search memory. This is mandatory for Standard+ work and skipped only for Tier 1.
Two searches minimum:
limit: 5.filter_tags: ["self-learning"],
limit: 5.The second search is the one people skip, and it is the one that pays. It is where "we tried that and it did not work" lives.
The MCP schema is strict, and the failure mode is a confusing error rather than a graceful degradation:
| Parameter | Type | Pass | Do not pass |
|---|---|---|---|
query | string | "cache eviction" | — |
limit | integer | 5 | "5" |
filter_tags | array of strings | ["self-learning"] | '["self-learning"]' |
tags | array of strings | ["gotcha", "redis"] | '["gotcha"]' |
When a learning moment happens, store it before proceeding. This is automatic, not optional — the correction you do not write down is the one you repeat in three weeks with a straight face.
| Trigger | Type |
|---|---|
| The user corrects your approach | correction — MANDATORY, never skip |
| You tried 3+ approaches before one worked | gotcha |
| Root cause turned out to be non-obvious | gotcha |
| You discovered an undocumented project convention | convention |
| An API or library behaved differently than expected | api-drift |
| A test failed for a non-obvious reason | error-fix |
[LRN:<type>]self-learning + the type + domain keywordsThe four-part body matters more than it looks. "Redis eviction is LRU here" is
a fact and will be forgotten. "We used LFU, it thrashed under burst traffic
because the counters never decayed, so this codebase pins LRU — check
cache.py before changing eviction" is a rule, and it survives.
The temptation is to finish the task and store the learning afterwards. There is no afterwards. The session ends, the context window closes, and the learning that existed only as an intention is gone. Store it while it is still true.
When --quality-locked is active, loop review/revise at each Quality Gate until findings reach a clean bar (no Critical, no Major, only cosmetic Minor) or the iteration cap (10) is reached. Uses a deterministic Python checker for classification and decision logic; agent runs the actual review and revision steps.
Use after implementation passes the quality gates, when a change needs an independent pass over the code before it ships, or when the user asks to "QA this", "check the implementation", or "verify it matches the spec". Applies to Standard and Comprehensive scopes with the Task tool available; Lightweight tasks skip.
Configure Sage preset and project settings. Switch between base, startup, enterprise, or opensource constitution presets. Use when the user says "configure sage", "change preset", or "sage settings".