القائمة
Save insights from completed tasks to role memory.
يُحدَّد Prompt الذي يراجع المصدر أولًا بشكل افتراضي. يمكنك التبديل إلى أمر مباشر أو تنزيل نسخة محلية.
اقرأ SKILL.md وأي ملفات مرافقة يعرضها SkillsMP قبل أن تقرر التثبيت.
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
يتجاوز الأمر المباشر Prompt المخصّص للمراجعة. افحص المصدر قبل تشغيله.
npx skills add https://github.com/Dwsy/agent --skill memory-retroيبقى الأمر في سطر واحد. مرّر أفقيًا لمراجعته كاملًا قبل النسخ.
تفضّل نسخة محلية؟ نزّل الملفات المتاحة حاليًا لدى SkillsMP.
Operate Glimpse-APPs with progressive context — app list in system, domain detail via tools on demand.
Load prior knowledge from role memory before starting any task.
Use when user asks to design, choose, explain, or implement UI transitions, motion patterns, easing/timing choices, state-change animations, shared-element effects, drill-down navigation, tab/filter/list/detail switching, or vague requests like "make this feel smoother", "more iOS", "more natural", or "animate this UI" that need concrete motion taxonomy and implementation guidance.
عرض SKILL.md
| name | memory-retro |
| description | Save insights from completed tasks to role memory. |
| whenToUse | After completing any task involving code changes, architectural decisions, debugging, or non-trivial problem solving. Distill what you learned into reusable memories so future sessions can benefit. Invoke proactively at task end — do not wait for the user to ask. |
You have access to a role-based persistent memory system. After completing this task, reflect on what you learned and save valuable insights.
| Tool | Purpose |
|---|---|
memory({ action: "add_learning", content: "..." }) | Save a durable insight. LLM auto-extracts tags. Auto-deduplicates. |
memory({ action: "add_preference", content: "...", category: "..." }) | Save a user preference. Goes directly to consolidated. |
memory({ action: "search", query: "..." }) | Check if similar memory already exists |
memory({ action: "reinforce", content: "..." }) | Increment [Nx] usage count for existing learning |
add_learning auto-deduplicates. Just call it — if similar text exists, it returns "Already stored". Then use reinforce instead.
memory({ action: "add_learning", content: "MyBatis-Plus getOne needs .last('LIMIT 1') to avoid TooManyResultsException" })
LLM auto-extracts tags (e.g., mybatis, gotcha). No manual tagging.
memory({ action: "add_preference", content: "用户偏好中文沟通，技术术语可保留英文", category: "Communication" })
memory({ action: "reinforce", content: "安全删除原则" })
→ "Reinforced [abc123] -> 6x"
Two paths to memory:
Auto-extract (agent_end, compaction)
→ pending.md [○] → search score ≥0.5 → auto-promote → consolidated [0x]
Manual (this tool)
→ consolidated [0x] directly
The pending layer filters noise. Only memories proven useful by actual usage survive. Manual entries skip pending because you're explicitly deciding they're worth keeping.
| Action | Effect | When |
|---|---|---|
reinforce | used+1 in consolidated | You used an existing memory |
| Search score ≥0.5 | auto-reinforce (used+1) | Found via search |
| Pending auto-promote | pending → consolidated | Pending entry is relevant |
consolidate | dedup + canonical rewrite | Routine maintenance |
| Category | When |
|---|---|
Communication | Language, style, tone |
Code | Style, conventions, abstraction |
Tools | CLI, editors, workflows |
Workflow | Process, review, deployment |
General | Everything else |
✅ Good:
memory({ action: "add_learning", content: "MyBatis-Plus getOne needs .last('LIMIT 1')" })
✅ Good (reinforce):
memory({ action: "reinforce", content: "安全删除原则" })
❌ Bad: "The user asked me to fix a bug" — too generic
❌ Bad: "Error at /src/index.ts:42" — copy-paste, no insight
add_learning auto-deduplicates — if similar exists, use reinforce instead.استنادا إلى تصنيف SOC المهني
