菜单
Triggered when the owner asks to improve your operations, behavior, or skills, or when performing a periodic self-reflection. Converts vague dissatisfaction or observed inefficiencies into concrete technical improvements (e.g., editing MEMORY.md, SOUL.md, USER.md, or custom skills).
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。
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/gal-Tab/sir_albert --skill self-reflection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。
Self-improving skills loop. Reads recent sessions (index + axcli analytics + transcripts), surfaces recommended skill edits — new trigger phrases, description tweaks, skill merges, new-skill candidates — and emits them as a draft for Gal to approve. Never auto-applies. Use when the user says "/retro", "retro", "self-improve skills", "what should the OS learn", "review my corrections", or "skill trigger fixes".
Decision memory: record a call once, auto-apply it every time the same question recurs — never re-decide what's already settled. Use when the user says "/decide", "record this decision", "we decided", "did we decide X before", "log this call", or "is this already decided".
The Discover/Think front door — a Core phase that picks the right thinking lens BEFORE a GTM, Building, or Automation task starts. Front-runs all three working modes; load it before committing to any approach or tool. Triggers: "help me think through X", "let's explore", "discover", "think this through", "which lens", "office hours", "WDYT", "where do I start with", "not sure how to approach", or any moment where choosing the right lens matters more than executing fast.
基于 SOC 职业分类
正在显示 SKILL.md
| name | self-reflection |
| description | Triggered when the owner asks to improve your operations, behavior, or skills, or when performing a periodic self-reflection. Converts vague dissatisfaction or observed inefficiencies into concrete technical improvements (e.g., editing MEMORY.md, SOUL.md, USER.md, or custom skills). |
This skill transforms the agent's behavior based on user feedback or self-audits. Inspired by continuous evolution concepts (like SkillClaw), this protocol ensures the agent doesn't just "promise to do better" but makes durable, concrete, technical changes to its own system files, memory, or skills.
Use this skill when:
"I'll be more careful next time" is NOT a valid outcome. Every reflection must result in a concrete technical change: modifying a prompt, updating MEMORY.md, altering a skill's SKILL.md, or creating a new cron job.
If the user's feedback is vague, ask 1-2 focused questions to pin down:
Audit ALL relevant parts of your system before proposing changes.
SOUL.md, MEMORY.md, USER.md, AGENTS.md.SKILL.md files in the skills/ directory.Rule: Read broadly, change surgically. Scan everything related, modify only what's needed.
Present your findings to the user for approval BEFORE executing changes. Structure your proposal:
MEMORY.md, SKILL.md, etc.) will be edited, and what text will be added/removed.After the user approves the plan, use your file editing tools (edit / write) or cron tool to make the actual changes.
MEMORY.md under a changelog or learned preferences section).sir_albert repo), commit the changes to preserve the evolution.Output should be in the user's preferred language (e.g., Hebrew).
🔍 אבחון הבעיה (Root Cause): [Brief explanation of why the agent acted incorrectly based on current system state]
🛠️ שינויים מוצעים (Proposed Changes):
⚠️ תופעות לוואי (Side Effects / Trade-offs): [Any potential risks or impacts on other tasks]
האם לאשר ולבצע את השינויים האלו בקבצי המערכת?