cogikyo/dotfiles/done
Claude Code and agent skill by cogikyo/dotfiles
- community
Checking that the SKILL.md file is still in the repository.
Use when the user invokes /done or asks to finish and close the current session; Collab runs comments, commits, and a qualifying live retro, then closes finished lanes and tags the session.
Install
npx vskill@latest install cogikyo/dotfiles/doneSKILL.md
Done
Collab runs this procedure attended in the current session, in the order below. Invoking /done approves the comment pass, a qualifying live retro, deleting this session tree's own scratch and stopping processes running from it, and closing finished lanes and the session. It approves nothing else; Git operations follow Collab's Git rules.
Comments
When this session changed code, send build/scribe a comment-only audit of exactly those changed hunks. Fix or delete stale comments; add one only where a reader would otherwise form a wrong belief. Skip when no code changed or a scribe pass already ran after the last edit.
Commit
Load commit and commit this session's reviewed work. Ask one question with the question tool only when ownership of a path or hunk is unclear. Leave other sessions' work dirty and skip when nothing is left.
Retro
Judge in-session, without subagents, using the Qualify bar in ../../improve/retro/SKILL.md over this session and its children. Drop candidates this session already fixed. When any candidate qualifies, load retro and run live mode now without asking first; its own confirmation exchange still applies. Commit approved retro changes under the Commit rules above. When nothing qualifies, say so in one line.
Knowledge base
Mail COGIKYO SCRIBE any cogikyo decisions or facts this session has not already sent, following cogikyo; skip when there are none.
Cleanup
Run cleanup.sql unchanged through sessions with ?1 = root session ID and ?2 = now − 3 days in epoch milliseconds. Select existing /tmp/opencode/<name> paths with shared = 0; keep shared = 1 paths and any of unclear ownership without asking (REFUSALS hides sessions from the shared check). Add ~/.local/state/opencode/scratch/<sessionID>/ for each session in the tree. Keep both scratch roots and the shared runs.jsonl. Keep any scratch or /tmp/opencode path that grep -r finds cited in /home/cullyn/cogikyo/docs/, including the containing directory selected for deletion. Record count and size, then stop processes whose working directory or arguments point into the selected paths. Delete with literal rm -rf -- /tmp/opencode/<name> … paths; send sandbox directories to trash -- <path>.
Close
Close without asking; /done already says the user wants the session closed. Ask once with the question tool only when closing looks like a mistake, such as own work left uncommitted, an open lane or pending decision, or a retro finding the user has not addressed; on no, stop without tagging. Load clear-lanes and close finished lanes under its judgment. Choose the state and flags below, then call tools.opencode.session_tag through execute, omitting sessionID.
Tags
Choose one state and any flags without asking; pass the names below in lowercase as state and flags. Pass all applicable flags, including existing ones that still apply, because the tool replaces them except for REFUSALS.
Pick the state by outcome:
CLOSED: the session's work landed.ARCHIVED: the session was shelved and nothing landed.RESEARCH: the session's result is findings rather than changes.
Add each flag that applies:
FIX: the work repaired a bug or regression.ISSUES: the session hit friction the retro did not resolve.REVIEWME: landed work deserves a later look, such as a risky or unreviewed change.TODO: follow-up work remains.REFUSALS: the session's content made a model stop or refuse; this excludes its history from agents'sessionsqueries.
Report
Report commit OIDs, retro outcome, scratch paths removed (count and size), paths kept as shared or unclear, dirty state left, and the tag written.
Security
- Security scanPassed, 90 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingconfig/opencode/skills/orchestration/done/SKILL.md
Popularity
- 7-day trend score
- 0
Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/cogikyo/dotfiles/done)