session-close
Skill6 starsChanged 27 days agoMITView on GitHub
Close the working session properly — write this session's episode, update progress.md ("ახლა" + log) and plan.md, record lessons and proposals, reconcile auto-summarised sessions, run the gates and commit. Use when the owner says the session is ending (e.g. "დავასრულოთ", "დავხუროთ", "ვამთავრებთ"), before /clear or a new terminal, or when the status line shows ↻3.
- Commits and pushes
22 other projects use a version of this.
Adopted by
Most recently taken up by
- kimkon1011/orgiast-claude-rulestheir own version, first seen Nov 13, 2026
- paureis/pau-skillstheir own version, first seen Nov 8, 2026
- pankaj-mahaur/Agent-Daemontheir own version, first seen Nov 6, 2026
- NathanForster/claude-commandstheir own version, first seen Nov 1, 2026
- ngmeyer/skillstheir own version, first seen Nov 1, 2026
- albertomarturelo/siitheir own version, first seen Nov 1, 2026
Used with
Projects with it also use session-open and handoff.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
vishnujchandranSkillsession-open
Restart-safe session starter that restores context, validates readiness, and creates a focused execution plan for the current development session.
Also in 4 projects that use this
MicrosoftSkillhandoff
Compact the current conversation into a handoff document for another agent to pick up.
Also in 3 projects that use this
Other versions22 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 22
- albertomarturelo/sii★ 19same name
- ngmeyer/skills★ 4same name
- paureis/pau-skills★ 2same name
- PUNCKe/manuscript-context-integrity★ 2same name
- toper2525ua-eng/claude-longrun★ 2same name
- vardan6/trackline★ 2same name
- AN0099/toothpaste-kit★ 1same name
- CorvusWing/CorvusWing★ 1same name
- farky-tech/containers★ 1same name
- NathanForster/claude-commands★ 1same name
12 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository10 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- adrSkill
- approveSkill
- feature-planSkill
- intentSkill
- observatorySkill
- orchestrateSkill
- reflectSkill
- specSkill
The file
We split this skill into the rules it gives an agent
3 rules. 0 are common, 1 is distinctive, and 2 are very specific to this project.
0
Common
Most of the 18 edited versions we read keep these.
1
Distinctive
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 13
Do every step. File content in Georgian prose.
- Very specific: line 37
Reconcile: every episode listed under needs_reconcile: read its auto summary, make sure progress.md and plan.md reflect it, then set …
- Very specific: line 56
If compacts ≥ 3, recommend a new terminal or /clear now that everything is saved.
“Common” is measured: kept by at least 6 in 10 of the 18 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
/session-close — distribute this session into the project's memory
Facts for this session (from .claude/hooks/episodic.py):
!bash .claude/hooks/py.sh .claude/hooks/episodic.py status ${CLAUDE_SESSION_ID}
Do every step. File content in Georgian prose; ids, paths and commands unchanged. Only facts that actually happened in this session — no plans presented as done.
- Episode (the path above; if it does not exist, create it with the frontmatter used in
memory/episodes/): fill## რა გაკეთდა,## გადაწყვეტილებები(with the reason),## ღია საკითხები,## შემდეგი ნაბიჯი,## წინადადებები; settitle(3–8 words). Add## აუდიტიwith the one line printed bybash .claude/hooks/py.sh tools/observe.py report --session ${CLAUDE_SESSION_ID} --brief(transcripts are deleted after ~30 days; this line is what git keeps). If it printsOBSERVE INCONCLUSIVE, write that. - progress.md — rewrite the block between
<!-- auto: current-state:start -->and:endas a dashboard of four short parts (ADR-0017): ფაზა და მიზანი · ბოლოს გაკეთდა (one line, with the episode link) · შემდეგი ნაბიჯი · მფლობელის მოლოდინში. The story goes into the log entry, never into the dashboard. Insert ONE entry at the top of theauto: logblock:### YYYY-MM-DD · [[EP-…]] · <title>+ 2–4 bullets. If an auto entry for this episode exists (<!-- ep:EP-…:start -->), replace it with yours. Never edit older entries. Each of the three auto blocks has its own cap (BLOCK_CAPSin.claude/hooks/episodic.py; a longer block is cut in the next session's brief): keep each under it. - plan.md — rewrite the
auto: current-phaseblock: the phase's goal and its open items, one line each; a finished item moves, as it was, to the phase's done list below the block. Add new backlog items. For an approved feature plan indocs/plans/: only tick boxes, move approved → in-progress → done, or append under## გადახრები(the guard allows exactly that). tasks/todo.md — tick finished items only with their receipt (✓ YYYY-MM-DD · <commit | მფლობელი | file>) and move them under## დასრულებული; rewrite theauto: todo-nowblock (ახლა / შემდეგი); add the concrete next tasks with the next freeD-NNN; delete done items older than 14 days. When the list outgrows the block's cap, the full task lines go under## რიგიbelow it and „შემდეგი" keeps each id with a short title, never as a- [ ]line (an id is used once).tasksgate must pass. - Reconcile every episode listed under
needs_reconcile: read its auto summary, make sure progress.md and plan.md reflect it, then setneeds_reconcile: falseandreconciled_at: <now_utc>in that episode (the only fields you may change in another session's episode). - Lessons — if something failed twice or cost real time, write a lesson in two places (ADR-0017). One line
in the index
memory/LESSONS.md:
and its story in- **L-NNN Title.** The rule. Guarded by: drill (`group`) | <another check> | none. [[L-NNN]]memory/lessons/L-NNN.md(frontmatterid,type: lesson,title,status: living; what happened and how we know the rule is right). The rule lives only in the index line. The index counts against the byte budget of what every session loads —bash .claude/hooks/py.sh tools/budget.pyshows what is left; a lesson that does not fit merges with or retires an old one (the guard, G7, refuses the growth). - Proposals — new skill / gate / ADR / spec / tool ideas go to
memory/PROPOSALS.mdas- [ ] **P-NNN** …(next free number) before<!-- auto: proposals:end -->. Never create a skill or install anything without the owner's yes. - Close the episode — set
closed_by: session-closeandclosed_at: <now_utc from above>. - Gates —
bash .claude/hooks/py.sh tools/gates/gate.py --pre-commit; fix every failure (never bypass, never weaken). - Commit — CHECKPOINT format (what was done / files / status), on the current branch.
- Tell the owner in 2–4 Georgian sentences: what was done, what is next, what waits for them.
If compacts ≥ 3, recommend a new terminal or
/clearnow that everything is saved. Then runbash .claude/hooks/py.sh .claude/hooks/episodic.py status ${CLAUDE_SESSION_ID}once more: unless itsengine:line is-, give that line to the owner word for word, on its own line — which engine files changed and whether their commit passed the drill (ADR-0015). Unless itsbackup:line is-, give that one word for word too, on its own line — commits that are only on this computer; upload (git push) only on their yes.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

