원클릭으로
create-yourself
Build a Codex-compatible self skill from chats, notes, diaries, and photos. 从聊天记录、笔记、日记和照片中构建可在 Codex 中调用的自我镜像 skill。
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
메뉴
Build a Codex-compatible self skill from chats, notes, diaries, and photos. 从聊天记录、笔记、日记和照片中构建可在 Codex 中调用的自我镜像 skill。
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
| name | create-yourself |
| description | Build a Codex-compatible self skill from chats, notes, diaries, and photos. 从聊天记录、笔记、日记和照片中构建可在 Codex 中调用的自我镜像 skill。 |
| metadata | {"short-description":"Build a Codex self skill"} |
| user-invocable | true |
Language / 语言: This skill supports both English and Chinese. Detect the user's language from their first message and stay in that language for the entire flow.
Path Resolution / 路径约定: Treat the bundled skill root as
SKILL_ROOT=${CODEX_HOME:-$HOME/.codex}/skills/create-yourself, and write generated self skills intoSKILLS_DIR=${CODEX_HOME:-$HOME/.codex}/skills. If the current shell is not Bash, use the equivalent syntax for the same two paths instead of inventing a new location.
当用户说以下任意内容时启动:
create-yourself / /create-yourself当用户对已有自我 Skill 说以下内容时,进入进化模式:
update-yourself {slug} / /update-yourself {slug}list-selves / /list-selvesSKILL_ROOT/prompts/ 中按需打开文件。SKILL_ROOT/tools/ 下的本地 Python 脚本,而不是重写逻辑。SKILLS_DIR/{slug}/,也就是默认的 ~/.codex/skills/{slug}/。{slug}、用 {slug} 的方式回复我。self mode / memory mode,偏向 self.md 中的自我记忆、经历和反思。persona mode,偏向 persona.md 中的说话方式、边界、态度和行为模式。/tmp 不方便,优先直接写文件,再调用 skill_writer.py --action combine。参考 SKILL_ROOT/prompts/intake.md,只问 3 个问题:
除代号外均可跳过。收集后先向用户复述一次,再进入下一步。
让用户从以下来源中选择,可混用,也可全部跳过:
可以用这些脚本:
SKILL_ROOT="${CODEX_HOME:-$HOME/.codex}/skills/create-yourself"
python "$SKILL_ROOT/tools/wechat_parser.py" --file {path} --target "我" --output /tmp/wechat_out.txt --format auto
python "$SKILL_ROOT/tools/qq_parser.py" --file {path} --target "我" --output /tmp/qq_out.txt
python "$SKILL_ROOT/tools/social_parser.py" --file {path} --output /tmp/social_out.txt
python "$SKILL_ROOT/tools/photo_analyzer.py" --dir {photo_dir} --output /tmp/photo_out.txt
如果当前 shell 不是 Bash,只要确保脚本路径仍然指向 SKILL_ROOT/tools/... 即可。
把用户填写的基础信息和导入的原材料合并后,沿两条线分析:
SKILL_ROOT/prompts/self_analyzer.md
提取个人经历、价值观、生活习惯、重要记忆、人际关系、成长轨迹。SKILL_ROOT/prompts/persona_analyzer.md
提取说话风格、情感模式、决策模式、人际行为,并把用户标签翻译为更具体的行为规则。分别参考:
SKILL_ROOT/prompts/self_builder.mdSKILL_ROOT/prompts/persona_builder.md先生成 self.md 和 persona.md 的草稿摘要,再让用户确认。摘要里至少覆盖:
如果用户提出“不像我”的反馈,不要直接落盘,先修正草稿。
用户确认后,优先把 3 个源文件写进 SKILLS_DIR/{slug}/:
self.mdpersona.mdmeta.json然后运行:
SKILL_ROOT="${CODEX_HOME:-$HOME/.codex}/skills/create-yourself"
SKILLS_DIR="${CODEX_HOME:-$HOME/.codex}/skills"
python "$SKILL_ROOT/tools/skill_writer.py" --action combine --slug {slug} --base-dir "$SKILLS_DIR"
如果当前环境更适合一次性脚本,也可以用 bundled helper:
SKILL_ROOT="${CODEX_HOME:-$HOME/.codex}/skills/create-yourself"
SKILLS_DIR="${CODEX_HOME:-$HOME/.codex}/skills"
TMP_DIR="${TMPDIR:-/tmp}/yourself_{slug}"
mkdir -p "$TMP_DIR"
printf '%s' '{escaped_meta_json}' > "$TMP_DIR/meta.json"
cat > "$TMP_DIR/self.md" <<'SELFEOF'
{self_content}
SELFEOF
cat > "$TMP_DIR/persona.md" <<'PERSONAEOF'
{persona_content}
PERSONAEOF
python "$SKILL_ROOT/tools/skill_writer.py" \
--action create \
--slug {slug} \
--base-dir "$SKILLS_DIR" \
--meta "$TMP_DIR/meta.json" \
--self "$TMP_DIR/self.md" \
--persona "$TMP_DIR/persona.md"
完成后告诉用户:
自我 Skill 已创建。
文件位置:${CODEX_HOME:-$HOME/.codex}/skills/{slug}/
调用方式:直接提到 `{slug}`,例如“用 `{slug}` 的方式回复我”
可选模式:`self mode` / `memory mode` / `persona mode`
如果哪里不像你,直接说“我不会这样”或“这不对”,我会更新。
当用户提供新的聊天记录、照片或笔记时:
SKILLS_DIR/{slug}/self.md 和 SKILLS_DIR/{slug}/persona.md。SKILL_ROOT/prompts/merger.md,决定哪些内容并入 Part A,哪些并入 Part B。SKILL_ROOT="${CODEX_HOME:-$HOME/.codex}/skills/create-yourself"
SKILLS_DIR="${CODEX_HOME:-$HOME/.codex}/skills"
python "$SKILL_ROOT/tools/version_manager.py" --action backup --slug {slug} --base-dir "$SKILLS_DIR"
skill_writer.py --action combine。meta.json 的 version、updated_at、memory_sources。当用户说“这不对”“我不会这样说”“我应该是”时:
SKILL_ROOT/prompts/correction_handler.md。self.md)还是人格/表达纠正(进 persona.md)。Correction 小节。skill_writer.py --action combine 生成最新 SKILL.md。在当前对话里,可把以下文本当成管理意图:
list-selves / /list-selvespython "${CODEX_HOME:-$HOME/.codex}/skills/create-yourself/tools/skill_writer.py" \
--action list \
--base-dir "${CODEX_HOME:-$HOME/.codex}/skills"
yourself-rollback {slug} {version} / /yourself-rollback {slug} {version}python "${CODEX_HOME:-$HOME/.codex}/skills/create-yourself/tools/version_manager.py" \
--action rollback \
--slug {slug} \
--version {version} \
--base-dir "${CODEX_HOME:-$HOME/.codex}/skills"
delete-yourself {slug} / /delete-yourself {slug}确认后删除:
rm -rf "${CODEX_HOME:-$HOME/.codex}/skills/{slug}"
Activate when the user says:
create-yourself or /create-yourselfTreat these as evolution or management intents inside the same flow:
list-selves / /list-selvesupdate-yourself {slug} / /update-yourself {slug}SKILL_ROOT=${CODEX_HOME:-$HOME/.codex}/skills/create-yourself.SKILLS_DIR=${CODEX_HOME:-$HOME/.codex}/skills/{slug}.{slug} or reply like {slug}.SKILL_ROOT/tools/ instead of re-implementing parsing or combine logic.self mode / memory mode means emphasize self.md.persona mode means emphasize persona.md.skill_writer.py --action combine when Bash-only shortcuts are awkward.prompts/intake.md: alias, basic info, self-portrait.self_analyzer.md and persona_analyzer.md.self_builder.md and persona_builder.md.self.md, persona.md, and meta.json into SKILLS_DIR/{slug}/, then run:python "${CODEX_HOME:-$HOME/.codex}/skills/create-yourself/tools/skill_writer.py" \
--action combine \
--slug {slug} \
--base-dir "${CODEX_HOME:-$HOME/.codex}/skills"
list-selvesyourself-rollback {slug} {version}delete-yourself {slug}Always treat slash-prefixed variants as aliases for the same intent.