菜单
Produce a portable bundle of the whole store — every entry as markdown, plus INDEX.md, scopes.yaml and a manifest. The conformance test for portability. Use when moving backends, backing up, or checking the store is actually portable.
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/danielrosehill/Claude-Personal-Context-Plugin --skill context-export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。
Assemble a scoped background briefing for a workspace or a task — read only the entries whose scopes match, and say which they were. Use at the start of a session in a consumer workspace, or when the user asks "what do you know about me for this".
First-run setup for the personal context layer — choose a backend, pick a store location, scaffold it, and write machine-local config. Run before any other personal-context skill. Triggers on "set up personal context", "where should my context live", "init personal context".
The write path — draft an entry in full, show it, confirm scope and sensitivity, write it, index it, commit it. Use when the user says "remember that", when another skill has candidates to write, or when a consumer workspace proposes writes back.
基于 SOC 职业分类
正在显示 SKILL.md
| name | context-export |
| description | Produce a portable bundle of the whole store — every entry as markdown, plus INDEX.md, scopes.yaml and a manifest. The conformance test for portability. Use when moving backends, backing up, or checking the store is actually portable. |
Implements §10 of the contract. Also the honesty test for the whole design: if the store cannot be exported in one command and read without this plugin, then it is a memory feature with extra steps.
Resolve paths (sh "${CLAUDE_PLUGIN_ROOT}/scripts/paths.sh" --json) and read the config. Ask
for a destination; default <store_root>/personal-context-export-YYYYMMDD/, i.e. alongside
the store rather than inside it.
Copy every entry as markdown, preserving the directory structure, plus INDEX.md and
scopes.yaml verbatim.
For backend: mcp, this is the real work: pull every entry from the server and render it to
the schema. If the server cannot round-trip the required frontmatter fields, say so plainly —
that backend does not conform, and the export is lossy in a way the person should know about
before they rely on it.
Write manifest.json at the export root:
{
"contract_version": 1,
"exported": "2026-08-14T12:00:00Z",
"backend": "filesystem",
"source_path": "<resolved store path>",
"entry_count": 34,
"types": {"person": 9, "period": 4, "event": 6, "pattern": 7, "care": 3, "preference": 5},
"scopes": ["general", "career", "family", "relationship", "health"],
"sensitivity": {"low": 4, "normal": 27, "high": 3}
}
Copy the contract itself alongside it (context-contract.md), so the bundle is
self-describing to whoever opens it in five years with no idea what produced it.
Verify: count files on disk against manifest.entry_count and against the lines in
INDEX.md. Three numbers that agree. If they don't, say which is authoritative (the files)
and what's inconsistent — a stale INDEX.md is the usual cause and is itself worth fixing.
Report the destination, the counts, and a one-line reminder that the bundle contains
everything including sensitivity: high entries.
If they ask for a partial export — one scope, or excluding sensitive entries — do it, and record
the filter in the manifest as "filter": {"scopes": ["career"], "include_sensitive": false}. A
filtered bundle must say it is filtered; an unlabelled partial export is how someone later
concludes material was lost.
Common reason for this: handing context to a different tool, or to a person. In that case offer
to strip sensitivity: high by default.