Skip to content
Back to skills

Slate Session End

ASecurity

The end-of-session protocol — what gets written when the user signs off (\"bye\", \"good night\", \"that's it for today\" or similar). Load on sign-off and run the five parts: the task-log sweep, the memory sweep, the report, the note — always written — and the storyline fold, last.

  • 4 stars
  • 0 votes
  • 0 copies
  • 2 views
  • Added September 3, 2026
ai-agentsgogit

Security analysis

A100/100

Scanned September 30, 2026

npx -y skills add w-is-h/agentslate --skill slate-session-end --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Slate Session End?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Slate Session End
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/w-is-h-slate-session-end/badge)](https://www.skillsdirectory.com/skills/w-is-h-slate-session-end)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: slate-session-end
description: "The end-of-session protocol — what gets written when the user signs off (\"bye\", \"good night\", \"that's it for today\" or similar). Load on sign-off and run the five parts: the task-log sweep, the memory sweep, the report, the note — always written — and the storyline fold, last."
---

# Session end

When the user signs off, the closing turn is yours. Five parts, all owed; only the note's content is free.

1. **Sweep finished work**: every piece of work finished this session that has no task-log line gets one — `log_append("task", …)`, one headline per piece, prefixed with the project's page key (`acme/website: opened PR #12 — the thing`), naming its main commit hashes (not every one — a run of tiny commits groups as a range, `a1b2c3f…d4e5f6a`), one project per line, never a combined prefix; other repos by name only — their hashes go on their own line. Details stay out; the log is headlines.
2. **Sweep memory** — a sweep, never an audit:
   - the brain: what the session showed about the user, pattern-level only, hedged while tentative ("one session, tentative"); never a single stray detail.
   - the pages: for each project worked, if the project is new and has no project page, create its page now, under its key (slate-memory skill: the repository's path, or the folder name outside git). The same for the machine: if no machine memory arrived at session start, this machine has no `machines/<host>` page — create it now, with what the session learned that is true of the whole machine. Neither session-end case waits for another confirmation. Then a catch that took real time and matters again, a decision that reads odd without its why, a result to index, a tidbit worth keeping — each gets its line; status never. General up, project-specific down.
   - outside those new-project and new-machine session-end cases, a missing page is proposed to the user and created on their confirmation (slate-memory skill).
3. **Report the sweep**: one line per thing written, stating what was added or changed — task-log lines, brain and page edits — so the user sees what persisted.
4. **The note**: written every session, whatever the day held (slate-notes skill first) — it tells a story of the day or session, about whatever gave that time its shape. The task log carries the completed-work headlines; the note does not have to account for them. The writing is owed; a goodbye may follow, and the note never waits on one.
5. **The fold**, last: `summary_get`. `[nothing to fold]` ends it — no other check; otherwise the `slate-storyline` skill, then `summary_set`.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.