gsd-mode
Skill0 starsChanged 38 days agoMITView on GitHub
Full blast day mode for Claude Code. No session limit stops, no handoff unless asked, only tasks that need no human decision (oldest first), one git worktree per agent, all questions batched into a single question list at the end. Persists for the day across sessions and projects. Triggers "/gsd", "/gsd on", "/gsd off", "full blast", "GSD mode", "blast day".
In this repository8 other files it gives its agents.
- autocomplete-miningSkill
- conversion-tracking-auditSkill
- image-processSkill
- search-term-reviewSkill
- secret-ledgerSkill
- seo-onpage-passSkill
- serp-teardownSkill
- ux-lawsSkill
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 2 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 12
No session limit stops. No "session at limit" warning, no cold start block, no handoff unless you ask.
- Very specific: line 14
Every agent session works in its own git worktree under .claude/worktrees/ and claims its tasks (for example | claim Name on the task line) …
- Distinctive: line 59
Never a handoff block, never a cold start block, never a session limit warning.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
GSD mode (get stuff done, full blast day)
Turn Claude Code into an uninterrupted working day with one command.
What the mode changes
- No session limit stops. No "session at limit" warning, no cold start block, no handoff unless you ask. You compact or clear when you decide.
- Agents work only tasks that need no human decision, oldest open tasks first, and never stop to ask mid batch; all questions go into a single question list at the end.
- Every agent session works in its own git worktree under
.claude/worktrees/and claims its tasks (for example| claim Nameon the task line) before the first tracked edit. - The mode persists across sessions for the whole day and across projects, read from
~/.claude/.gsd-mode.
These override your project rules on session limits, handoff timing and mid task check ins for the day. Approval rules stay: deploys, live backend writes, page deletes, URL changes and pushes that trigger CI still need your approval; they are queued into the final question list.
Commands
/gsd on: write today's date to the flag:date +%F > ~/.claude/.gsd-mode/gsd off: delete the flag:rm -f ~/.claude/.gsd-mode/gsd: status:cat ~/.claude/.gsd-mode 2>/dev/null; ON when the date equals today, else OFF.
Hook
scripts/gsd-mode-hook.mjs is a UserPromptSubmit hook. While the flag date equals today it injects the mode line into every prompt; a stale flag is deleted silently. Install:
mkdir -p ~/.claude/hooks
cp ~/.claude/skills/gsd-mode/scripts/gsd-mode-hook.mjs ~/.claude/hooks/Then add to ~/.claude/settings.json:
{
"hooks": {
"UserPromptSubmit": [
{ "hooks": [ { "type": "command", "command": "node ~/.claude/hooks/gsd-mode-hook.mjs" } ] }
]
}
}Task picking rule
- Open tasks (
- [ ]) in your project task list (for exampleTODO.md). - Skip any task marked as needing a human decision.
- Skip any task whose linked spec still has an open question for you.
- Skip tasks already claimed by another live agent.
- Sort by task date, oldest first. Work in that order.
Reply format in mode
- Short progress lines while working are allowed.
- One joint reply per batch, after every agent and check has finished.
- Questions only at batch end, as a single numbered question list, recommended option marked.
- Never a handoff block, never a cold start block, never a session limit warning.
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.

