From claude-stack
Quick mode for small tasks - compressed discuss+plan+build workflow without full ceremony
How this skill is triggered — by the user, by Claude, or both
Slash command
/claude-stack:s-quickThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Compressed workflow for small tasks. Skips ceremony, keeps discipline.
Compressed workflow for small tasks. Skips ceremony, keeps discipline.
Usage: /s:quick "description of the task"
Flags: --discuss | --full | --research
Accept the task description as the first argument. Parse optional flags:
| Flag | Effect |
|---|---|
| (none) | Compressed workflow - fastest path |
--discuss | Add full requirements gathering phase |
--full | Redirect to full ceremony: /s:new > /s:discuss > /s:plan > /s:build |
--research | Research the topic first using Context7 and web search, then implement |
If --full flag is present, inform the user and hand off to the full chain:
[QUICK] Full ceremony requested. Redirecting to /s:discuss > /s:plan > /s:build chain.
Then stop - let the full skills take over.
Unless --discuss flag is set, keep this MINIMAL:
Ask at most 3 clarifying questions. Prefer yes/no or multiple choice:
[QUICK] Task: "{user_description}"
Quick check:
1. Which files/modules does this affect? (or "auto-detect")
2. Any constraints I should know about? (or "none")
3. Should I create tests? (yes/no, default: yes)
If --discuss flag is set, run the full /s:discuss protocol instead.
If --research flag is present:
Display:
[QUICK] Researching: {topic}
[QUICK] Found: {N} relevant docs, {M} existing patterns
[QUICK] Proceeding with research context loaded.
NO separate plan file. Build a mental model and display it:
[QUICK] Plan:
1. {step 1} → {file}
2. {step 2} → {file}
3. {step 3} → {file}
Tests: {test file(s)}
Estimated changes: {N} files
This plan is NOT saved to docs/plans/. It exists only in this conversation.
Exception: if the plan exceeds 5 steps, suggest upgrading to full mode:
[QUICK] This task has {N} steps. Consider using /s:plan for a tracked plan.
Continue in quick mode? (yes/no)
TDD is NON-NEGOTIABLE even in quick mode. The cycle is lighter but still enforced:
Display after each phase:
[QUICK] RED: test written, fails as expected
[QUICK] GREEN: implementation done, test passes
[QUICK] REFACTOR: cleaned up, all tests pass
Lighter than full /s:verify but still evidence-based:
Display:
[QUICK] Verification:
New tests: PASS ({N} assertions)
Existing tests: PASS ({M} total)
Files changed: {list}
Update .planning/STATE.md with a lightweight entry:
## Quick Task: {description}
**Date:** {YYYY-MM-DD}
**Status:** complete
**Files:** {list of changed files}
**Tests:** {test file(s)}
This is a shorter format than full skill updates but still maintains the project memory.
Based on what was done:
[QUICK] Done. Options:
- /s:review → Get a code review
- /s:ship → Push and create PR
- /s:compound → Save learnings (recommended if you learned something new)
/s:plan > /s:build/s:brainstorm > /s:plan/s:auto/s:discuss firstany types)npx claudepluginhub bdarbaz/claude-stack-pluginExecutes ad-hoc tasks via spawned executor subagent with atomic commits and .planning/quick/ tracking. For small self-contained work, skips full plan/review.
Executes small, self-contained tasks like bug fixes, config changes, or minor refactors without full workflow overhead and with atomic commits.