跳至正文
OpenAgentSkill
浏览已开放 · 目录记录日期 2026-11-21 · 自动更新和账户服务暂时关闭。
搜索与研究

end

Close a work session - asks what was completed, what was learned and what comes next, then updates dashboard, todo, learnings, errors and writes memory/daily/YYYY-MM-DD.md. Use when user says "end session", "wrap up", "I'm done for today", "save progress", "let's close", "log this session". Do NOT use for a week-level retrospective (use /weekly-review) or for cross-session pattern mining (use /reflect).

onurpolat05 · 已有静态检查记录

安装与使用

运行命令前先阅读来源与依赖说明。复制命令不会自动安装或执行。

npx skills add onurpolat05/ALBA --skill end

技能说明


name: end description: Close a work session - asks what was completed, what was learned and what comes next, then updates dashboard, todo, learnings, errors and writes memory/daily/YYYY-MM-DD.md. Use when user says "end session", "wrap up", "I'm done for today", "save progress", "let's close", "log this session". Do NOT use for a week-level retrospective (use /weekly-review) or for cross-session pattern mining (use /reflect). context: inline effort: medium allowed-tools: [Read, Write, Edit, Glob, AskUserQuestion]

/end - Session End

Close the session with a structured 3-question protocol. Save state for next session.

Steps

1. Ask 3 Questions (one at a time, keep it short)

Q1: "What was completed this session?" (max 3 bullets) Q2: "Any learnings or insights?" (optional - user can skip) Q3: "What's the priority for next session?" (1 sentence)

2. Auto-Update (after answers)

Dashboard (memory/state/dashboard.md):

  • Mark completed items as [x]
  • Add next priority if mentioned
  • Update any status changes

Todo (memory/state/todo.md):

  • Mark completed tasks as [x]

Learnings (memory/knowledge/learnings.md):

  • If Q2 had content, append in format:
## [Topic] - YYYY-MM-DD
**Context:** [What prompted this learning]
**Insight:** [The learning itself]
**Application:** [When to apply this]

Errors (memory/knowledge/errors.md):

  • Check memory/knowledge/errors_raw.log for new entries from this session
  • If errors were resolved during the session, consolidate them into errors.md:
## [Error Type] - YYYY-MM-DD
**Pattern:** [What triggers this error]
**Solution:** [How it was resolved]
**Prevention:** [How to avoid it]

Daily Log (memory/daily/YYYY-MM-DD.md):

  • Create with session summary:
# YYYY-MM-DD

## Completed
- [From Q1 answers]

## Learnings
- [From Q2, if any]

## Next
- [From Q3]

3. Confirm

Output a 3-line summary:

Session saved.
Completed: [count] items | Learnings: [count]
Next: [priority from Q3]

Rules

  • Keep questions SHORT - this is CLI, not a form
  • If user says "nothing" or skips Q2, that's fine - don't push
  • Don't read the entire dashboard - just update what changed
  • If daily log already exists (multiple sessions per day), append to it
  • Create memory/daily/ directory if it doesn't exist
  • Never fail silently - confirm what was saved
  • Respect user's language (follow dashboard/CLAUDE.md language)

Note on Auto-Memory vs ALBA Memory

Claude Code's auto-memory system writes to ~/.claude/projects/<encoded-project-path>/memory/MEMORY.md. That index is capped at 25KB / 200 lines — entries past the cap are truncated, silently. Keep auto-memory entries to one line (~150 chars); the content lives in the linked file, not in the index.

ALBA's own memory/ (state, knowledge, projects, daily) is separate — no size cap, but keep entries short for readability. Auto-memory and ALBA memory coexist (see memory-compatibility.md).