session-end
Skill0 starsChanged 6 days agoMITView on GitHub
Wraps up a session — optional handoff, skill-update proposals from repeated errors, a PR audit. Use when the user signals the job is done. Not for mid-job checkpoints.
30 other projects use a version of this.
Adopted by
Most recently taken up by
- timursafiullin/aodtheir own version, first seen Nov 15, 2026
- mikkiola/radartheir own version, first seen Nov 15, 2026
- spizganed/skillstheir own version, first seen Nov 8, 2026
- edxleon/runspectortheir own version, first seen Nov 8, 2026
- salvo8519/fairmarktheir own version, first seen Nov 8, 2026
- puritysb/AgentDecktheir own version, first seen Nov 2, 2026
Used with
Projects with it also use session-start.
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.
bitwize-music-studioSkillsession-start
Runs the session startup procedure - verifies setup, loads config and state, checks skill models, and reports project status. Use at the beginning of a fresh session.
Also in 12 projects that use this
Other versions30 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 30
- puritysb/AgentDeck★ 299same name
- faizkhairi/claude-code-blueprint★ 73same name
- Kanevry/session-orchestrator★ 53same name
- valpere/session-indexer★ 42same name
- namedtoaster/The-Last-Mile★ 39same name
- iamneilroberts/claude-skills★ 26same name
- IlyaGorsky/memory-toolkit★ 13same name
- edxleon/runspector★ 11same name
- mae616/ai-template★ 9same name
- freskhu/genesis★ 5same name
20 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 repository22 other files it gives its agents.
- AGENTS.mdAGENTS.md
- claude-arsenal/AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- coverage-gapsSkill
- dep-upgradeSkill
- designSkill
- executionSkill
- explore-ideaSkill
- gate-checkSkill
- githubSkill
- initSkill
- litcriticSkill
- mutmut-reportSkill
- pin-checkSkill
- pypi-releaseSkill
- python-bootstrapSkill
- queue-addSkill
- queue-nextSkill
- queue-statusSkill
- reviewSkill
- shipSkill
- specifySkill
The file
We split this skill into the rules it gives an agent
3 rules. 0 are common, 0 are distinctive, and 3 are very specific to this project.
0
Common
Most of the 23 edited versions we read keep these.
0
Distinctive
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 43
When the mode is yes, load handoff-mode for the template, then:
- Very specific: line 72
Before the audit, render the review digest and hand it over in this reply (claude-arsenal:core:init § references/digest.md § 2), so what …
- Very specific: line 116
When arsenal/project/<WORKSPACE>/ exists, Step 1 writes arsenal/project/<WORKSPACE>/handover.md and refreshes arsenal/session/handover.md …
“Common” is measured: kept by at least 6 in 10 of the 23 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-end
End-of-job wrap-up in three steps: an opt-in handoff, a retrospective that proposes skill updates from recurring friction, and a report on the session's PRs.
CANARY: session-end-loaded-2026-05-20-4896c0a5-8ca7505c91dc34e6
When to load
- The user types
/session-endor says "wrap up", "we're done", "close this session". - The opt-in auto-fire
SessionStarthook runs (about once a week). - The github skill is about to open a PR and the handoff marker is
yes, sostatus/handoff.mdis regenerated before the PR commit.
Mid-job is too early: the retrospective needs a complete arc to scan.
Step 1 — handoff (opt-in)
First, if this session adopted or locked an architecture decision, make sure the
spec and plan say so (arsenal/project/<WORKSPACE>/spec.md + plan.md, or
status/specification.md + status/plan.md), or seed a queue task for it. A
blocking spec divergence found this session becomes a queue task too. An idea or
decision raised this session but missing from status/ideas.md is a handover item.
The handoff is a snapshot the next session overwrites; the spec, plan and queue are
the ledger.
Then read the host repo's CLAUDE.md for the marker:
| Marker | Behavior |
<!-- session-end: handoff=yes --> | Generate status/handoff.md from the session, commit it. |
<!-- session-end: handoff=ticket --> | Skip (one session = one ticket; the PR description suffices). |
<!-- session-end: handoff=no --> | Skip (the project does not use handoffs). |
| (no marker) | Ask the user once which mode this repo uses, then write the marker. |
When the mode is yes, load handoff-mode for the template, then:
python3 "${CLAUDE_SKILL_DIR}/scripts/create_handoff.py" --output status/handoff.mdThe script renders the template; fill in the session-specific content (done, to do, repro, do-not-touch) from the conversation, then write the file.
Step 2 — retrospective
python3 "${CLAUDE_SKILL_DIR}/scripts/query_session_history.py" --days 7 --limit 10The JSON report lists mechanical signals: repeated tool errors, throwaway scripts in
tmp/, repeated user corrections, repeated failing commands. Judge which are
recurring friction rather than normal noise, and offer the user a short list of
skill-update proposals. Once they accept one, load
retrospective-rubric for the block format and
append each accepted proposal to:
| Where the session runs | Target file |
This marketplace repo (plugins/<plugin>/skills/<skill>/ exists) | plugins/<plugin>/skills/<skill>/IMPROVEMENTS.md |
| Anywhere else | ~/.claude/proposed-skill-improvements/<YYYY-MM-DD>.md |
Step 3 — PR audit (when a queue exists)
Before the audit, render the review digest and hand it over in this reply
(claude-arsenal:core:init § references/digest.md § 2), so what waits on the user
reaches them even after a long unattended run.
This step reports and repairs nothing: merged task PRs already closed their tasks, and the queue workflow releases abandoned claims, so a session that ends abruptly still leaves a correct queue.
Task issues carry no PR link, so map open PRs to tasks by branch:
open_task_pr.sh names each arsenal/<task-id>-<slug>. Take the open PRs on such
branches, plus every claimed task (no PR yet shows as escalated or in progress),
and check each PR:
gh pr list --state open --json number,headRefName --jq '.[] | select(.headRefName | startswith("arsenal/"))'gh pr view <pr-url> --json title,state,mergeable,reviewDecision,statusCheckRollup \
--jq '{title,state,mergeable,reviewDecision,ci:([.statusCheckRollup[]?|.conclusion]|unique)}'Print a table for the user to review before the session closes:
| Task | PR | CI | Reviews | Mergeable | Action needed |
| t-a3f8 | #NNN | ✓ passing | approved | yes | — |
| t-b2c1 | #NNN | pending | — | CONFLICTING | Rebase required |
Mark a PR BLOCKED when CI fails, a review requests changes, or the branch
conflicts. Without gh, list task id, PR URL and title for a manual check. Always
list escalated tasks (retry cap exhausted, no PR) separately: they need a human
reset, not a PR review.
Auto-fire (opt-in)
A detached SessionStart hook can run this skill about once a week without an
explicit /session-end. The user installs it through the update-config skill; this
skill never edits settings.json itself. Load
auto-fire-setup when the user wants to wire it up,
skip the next run, or asks why an installed auto-fire is not firing.
Workspace-aware paths
When arsenal/project/<WORKSPACE>/ exists, Step 1 writes
arsenal/project/<WORKSPACE>/handover.md and refreshes arsenal/session/handover.md
instead of status/handoff.md. The marker still decides whether Step 1 runs.
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.

