跳到正文
cccskills
免费在 GitHub 公开

context-capsule

Write a resume-ready capsule of the current session (objective, what it is waiting on, the next concrete action) so a /clear or an auto-compact can pick up exactly where this left off. Write it before compaction; nothing writes one for you at that point. Use for a checkpoint before risky work, before /clear, or when a thread ships.

中文简介正在准备中。当前显示原文说明。

wrg32786/aigent-os星标 25(GitHub 25 + cccskills 0)

实测评分 未实测

查看安装方法

包含的文件(1)

  • SKILL.md3.6 KB

SKILL.md(原文)

安装前可以查看交给 AI 代理的指令内容。

context-capsule

The whole job: reconcile, write the capsule, confirm it, then stop.

A capsule is a small markdown file with YAML frontmatter. A fresh session reads three fields from it and continues. It is state for the next session, not a transcript of this one.

Steps

  1. Reconcile from live state. Look at what is actually true now: git status, git log -n 5, the files you changed, any task list. Budget: 2 to 4 reads. Record what happened, not what was said. A capsule that restates the conversation instead of the repo state is the failure this skill exists to prevent.

  2. Write it with the helper, which stamps the date, validates the fields and writes into the configured capsule directory:

    node "${CLAUDE_PLUGIN_ROOT}/scripts/capsule-lib.mjs" write \
      --objective "<what this work is for, one sentence>" \
      --waiting-on "<concrete items, owners, gates a fresh session must know>" \
      --next-action "<the one next step, specific enough to act on cold>" \
      --slug "<short-topic>" \
      --dir '${user_config.capsuleDir}'
    

    Run it from the project directory (do not cd elsewhere) and keep the --dir line exactly as written, single quotes included (they stop the shell expanding the placeholder when the option is unset). A configured path that itself contains a single quote is not supported. It carries the configured capsule directory. If the directory option is not set, that value is empty or still reads ${user_config.capsuleDir} literally; then drop the --dir line and the helper uses the project default.

    Optional: --body-file <path> with your own body sections (see Format).

  3. Confirm. The helper prints the path it wrote. Run node "${CLAUDE_PLUGIN_ROOT}/scripts/capsule-lib.mjs" validate <that path> and expect VALID. If it prints INVALID, fix the fields and write again.

  4. Stop. Tell the user the capsule path in one line. If they asked for /refresh, tell them to run /clear next.

Field rules

  • objective, waiting_on, next_valid_action are required and must be non-empty. waiting_on is the resume contract: write it so a fresh session can act from it alone.
  • next_valid_action must be a real action, not "resume the session" or "re-read the state". The helper refuses ceremony text.
  • Do not paste secrets, tokens or credentials into a capsule. It is a plain file in the project.

Format

Frontmatter fields: id, parent_capsule_id, status: active, objective, waiting_on, resume_trigger, expires, trigger, next_valid_action, success_criteria, tags, created_at (ISO-8601 with offset), resolved_at. Body sections: Done (don't redo), Historical-Errors -> Resolutions, Historical-Rejected-Approaches, Files-Read / Files-Modified, Operating-Facts, Pending-Gates, Claimed-Rows. The format is compatible with the full aigent-os install.

Lifecycle

active then resumed (set automatically by the SessionStart hook when a /clear loads it) then done. Never edit status by hand. Capsules older than 14 days (plugin option maxAgeDays, or the CONTEXT_CAPSULE_MAX_AGE_DAYS env var, to change) are refused at resume with a message, not loaded.

When not to capsule

Mid-thought (finish the thought first), or a trivial session a fresh start can rebuild from the repo.

评价

还没有评价。欢迎分享你的使用感受。

评论

无需登录即可发表。欢迎分享使用感受、提问,或说明在哪些环境下无法运行。

还没有评论。来写第一条吧。

提交后会立即公开。请勿填写邮箱、电话等个人信息。URL 最多 2 个。

0 / 1000 字

此表单受 reCAPTCHA 保护,适用 Google 的隐私权政策和服务条款。

同一仓库的技能

Track and surface sub-agent dispatch calibration ratios. Extract from JSONL transcript or read existing ledger. Per-agent clean/blocked/errored/partial trends.

中文简介正在准备中。当前显示原文说明。

wrg32786/aigent-os252026年10月29日 更新

Compose the AIgent's body state by reading existing signals. Lazy compute. Never polled. Reports the five pressures + recommended reflex.

中文简介正在准备中。当前显示原文说明。

wrg32786/aigent-os252026年10月29日 更新

brief

免费

Generate a structured delegation brief for any task

中文简介正在准备中。当前显示原文说明。

wrg32786/aigent-os252026年10月29日 更新

Quarterly skill that walks the skill-index against the actual SKILL.md files on disk, reports drift (skills indexed that no longer exist, skills on disk that aren't enrolled, skills whose descriptions have changed since enrollment). Hardens the auto-enrollment pipeline against silent rot.

中文简介正在准备中。当前显示原文说明。

wrg32786/aigent-os252026年10月29日 更新

When a [CADDY] hint fires and the principal is confused why, /caddy-explain walks the scoring transparently — which trigger matched, what points it earned, what the threshold was, what other skills were close. Pairs with the legibility thesis. Caddy's decisions should be inspectable on demand.

中文简介正在准备中。当前显示原文说明。

wrg32786/aigent-os252026年10月29日 更新

Enroll a skill into Caddy's index so it can be surfaced when the right prompt arrives. Use when a new skill is created, copied in, or when a skill needs updated triggers. Usage — "/caddy-enroll <path to SKILL.md>".

中文简介正在准备中。当前显示原文说明。

wrg32786/aigent-os252026年10月29日 更新

查看 wrg32786 的全部技能

报告此技能的问题