How do I install the Gsd Loop Schedule skill?

Run npx skillmds@latest add open-gsd/gsd-loop-schedule in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Gsd Loop Schedule safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Gsd Loop Schedule?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Gsd Loop Schedule free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Gsd Loop Schedule?

open-gsd (@open-gsd) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Gsd Loop Schedule by open-gsd — Agent Skill

Name: Gsd Loop Schedule
Author: open-gsd

Gsd Loop Schedule

Create or update a native recurring task that repeatedly runs one-pass gsd-loop builder or reviewer work. Use when asked to keep the build or review queue running, schedule gsd-loop, or automate repeated passes.

open-gsd 611620a 2 files · 3.0 KB Updated Sep 17, 2026

File contents

Use the host's native recurring-task tool. In Codex, use the native scheduled-task tool. If the host has no recurring-task capability, stop and explain that this scheduling skill is unsupported there. Do not start a shell while loop or launch a separate agent process.

Resolve the repository root and owner/repo. Schedule exactly one lane in the current chat: build by default, or review when explicitly requested.
Run npx @opengsd/gsd-loop@latest doctor before build scheduling. Add --review-ready before review scheduling. Do not create or update a review task when that stricter check fails.
Use the deterministic name gsd-loop LANE — owner/repo. Inspect existing scheduled tasks first. Update an exact match instead of duplicating it. If a different active builder targets the same repository, stop and identify it.
Attach the task to the current chat, start at 15 minutes, and initialize its lane-specific idle count to zero. Keep any repository-local scheduler state at .gsd/scheduled_tasks.lock; never write it under a harness-specific directory such as .claude/.
Put the lane skill in the scheduled prompt using the current host's native invocation syntax:
- Codex: $gsd-loop-build or $gsd-loop-review
- Claude Code: /gsd-loop-build or /gsd-loop-review
- Cursor: /gsd-loop-build or /gsd-loop-review
- Gemini CLI: Use the gsd-loop-build skill or Use the gsd-loop-review skill
- Grok Build: /gsd-loop-build or /gsd-loop-review
- Kimi Code: /skill:gsd-loop-build or /skill:gsd-loop-review Run exactly one playbook pass per wake and require its final GSD_LOOP_RESULT line. A missing or malformed result is blocked, never inferred from prose.
After each pass, run npx @opengsd/gsd-loop@latest policy EVENT IDLE_COUNT. Persist the returned count in task context and immediately apply its action and interval to this scheduled task. This resets productive work to 15 minutes, backs idle work off to 60 minutes, and pauses after three idle passes.
Treat credentials, permissions, dirty unrelated worktrees, escalations, and other playbook stop conditions as blocked; pause and report them.

Build and review require separate chats and separate scheduled tasks so their state cannot mix. If the user requests both, schedule the current lane only and tell them to invoke this skill in a second chat for the other lane.

Never schedule the interactive discover or spec lanes. Never merge, enable auto-merge, force-push, or run more than one builder for the same repository.

---name: gsd-loop-scheduledescription: Create or update a native recurring task that repeatedly runs one-pass gsd-loop builder or reviewer work. Use when asked to keep the build or review queue running, schedule gsd-loop, or automate repeated passes.---Use the host's native recurring-task tool. In Codex, use the nativescheduled-task tool. If the host has no recurring-task capability, stop andexplain that this scheduling skill is unsupported there. Do not start a shell`while` loop or launch a separate agent process.1. Resolve the repository root and `owner/repo`. Schedule exactly one lane in the current chat: build by default, or review when explicitly requested.2. Run `npx @opengsd/gsd-loop@latest doctor` before build scheduling. Add `--review-ready` before review scheduling. Do not create or update a review task when that stricter check fails.3. Use the deterministic name `gsd-loop LANE — owner/repo`. Inspect existing scheduled tasks first. Update an exact match instead of duplicating it. If a different active builder targets the same repository, stop and identify it.4. Attach the task to the current chat, start at 15 minutes, and initialize its lane-specific idle count to zero. Keep any repository-local scheduler state at `.gsd/scheduled_tasks.lock`; never write it under a harness-specific directory such as `.claude/`.5. Put the lane skill in the scheduled prompt using the current host's native invocation syntax: - Codex: `$gsd-loop-build` or `$gsd-loop-review` - Claude Code: `/gsd-loop-build` or `/gsd-loop-review` - Cursor: `/gsd-loop-build` or `/gsd-loop-review` - Gemini CLI: `Use the gsd-loop-build skill` or `Use the gsd-loop-review skill` - Grok Build: `/gsd-loop-build` or `/gsd-loop-review` - Kimi Code: `/skill:gsd-loop-build` or `/skill:gsd-loop-review` Run exactly one playbook pass per wake and require its final `GSD_LOOP_RESULT` line. A missing or malformed result is `blocked`, never inferred from prose.6. After each pass, run `npx @opengsd/gsd-loop@latest policy EVENT IDLE_COUNT`. Persist the returned count in task context and immediately apply its action and interval to this scheduled task. This resets productive work to 15 minutes, backs idle work off to 60 minutes, and pauses after three idle passes.7. Treat credentials, permissions, dirty unrelated worktrees, escalations, and other playbook stop conditions as `blocked`; pause and report them.Build and review require separate chats and separate scheduled tasks so theirstate cannot mix. If the user requests both, schedule the current lane only andtell them to invoke this skill in a second chat for the other lane.Never schedule the interactive discover or spec lanes. Never merge, enableauto-merge, force-push, or run more than one builder for the same repository.

Frequently asked questions