Help us improve
Share bugs, ideas, or general feedback.
From lighterpowers
Use when you have a comprehensive, well-researched plan ready for execution without oversight during task implementation
npx claudepluginhub bradwindy/lighterpowersHow this skill is triggered — by the user, by Claude, or both
Slash command
/lighterpowers:one-shot-developmentThis skill is limited to the following tools:
The summary Claude sees in its skill listing — used to decide when to auto-load this skill
Execute all plan tasks sequentially without human checkpoints, then run build and test phases with fix loops.
Guides technical evaluation of code review feedback: read fully, restate for understanding, verify against codebase, respond with reasoning or pushback before implementing.
Share bugs, ideas, or general feedback.
Execute all plan tasks sequentially without human checkpoints, then run build and test phases with fix loops.
Core principle: Trust the plan. Execute everything, validate at the end.
Warning: This mode is NOT recommended for most use cases. Use only when:
Use this skill when:
Don't use when:
docs/lighterpowers/plans/feature.md)Load Plan -> Phase 1: Pre-Execution Setup
-> Phase 2: Implementation (all tasks, no checkpoints)
-> Phase 3: Build Phase (max 3 fix cycles)
-> Phase 4: Build->Test Checkpoint (user decides)
-> Phase 5: Test Phase (max 3 fix cycles)
-> Phase 6: Completion
Same as other execution modes. Present offers before task loop begins.
### Pre-Execution VerificationBefore starting implementation:
User can decline any offer - the requirement is presentation, not execution.
Check if on main/master/develop:
git branch --show-current
If on base branch:
AskUserQuestion(
questions: [{
question: "Create feature branch for this work?",
header: "Branch",
options: [
{label: "Yes", description: "Create branch: feature/<plan-name>"},
{label: "Skip", description: "Stay on current branch"}
],
multiSelect: false
}]
)
If primary issue identified (from plan header or branch name):
AskUserQuestion(
questions: [{
question: "Update issue [ISSUE-ID] status to in-progress?",
header: "Status",
options: [
{label: "Yes", description: "Mark issue [ISSUE-ID] as in-progress"},
{label: "Skip", description: "Leave status unchanged"}
],
multiSelect: false
}]
)
After offers complete (or skipped), proceed to Phase 2.
Purpose: Execute ALL tasks from the plan sequentially without pausing for human input.
Update docs/lighterpowers/current-progress.md after each task:
# One-Shot Development Progress
## Plan
docs/lighterpowers/plans/feature-plan.md
## Mode
One-Shot (no checkpoints)
## Status
Implementing Task 4/10
## Completed Tasks
- [x] Task 1: Setup project structure
- [x] Task 2: Add base API client
- [x] Task 3: Add error handling
## Remaining Tasks
- [ ] Task 4: Add retry logic
- [ ] Task 5: Add rate limiting
...
## Discovered Work
- [ ] "Need timeout configuration" (discovered in Task 3)
**Implementation Completion Gate** (Required):
STOP CONDITION: If ANY unchecked, do NOT proceed. Complete all tasks before proceeding to Phase 3.
Purpose: Validate that all implemented code compiles/builds successfully.
Detect build command from project manifests (in priority order):
| Manifest File | Build Command |
|---|---|
package.json (with scripts.build) | npm run build |
Cargo.toml | cargo build |
pyproject.toml | python -m build |
Makefile | make |
go.mod | go build ./... |
If no manifest found:
AskUserQuestion(
questions: [{
question: "No build system detected. What build command should I use?",
header: "Build",
options: [
{label: "Skip build", description: "No build needed for this project"},
{label: "Custom command", description: "I'll provide a build command"}
],
multiSelect: false
}]
)
fix_cycles = 0
MAX_CYCLES = 3
while fix_cycles < MAX_CYCLES:
run build command
if build succeeds:
proceed to Phase 4
break
if build fails:
analyze error output
apply targeted fix
fix_cycles += 1
if fix_cycles >= MAX_CYCLES:
escalate to user
After 3 failed fix cycles:
AskUserQuestion(
questions: [{
question: "Build failed after 3 fix attempts. How do you want to proceed?",
header: "Escalate",
options: [
{label: "Continue trying", description: "Reset counter and try more fixes"},
{label: "Skip build", description: "Proceed to checkpoint without passing build"},
{label: "Stop", description: "Pause execution, I'll investigate"}
],
multiSelect: false
}]
)
**Build Phase Gate** (Required):
STOP CONDITION: If ANY unchecked, do NOT proceed.
Purpose: The ONLY human checkpoint in one-shot mode.
AskUserQuestion(
questions: [{
question: "Build successful. How would you like to proceed?",
header: "Continue",
options: [
{label: "Run tests", description: "Continue to test phase with fix loops"},
{label: "Review first", description: "Pause here so I can review the changes"},
{label: "Stop here", description: "Skip tests, proceed to completion"}
],
multiSelect: false
}]
)
STOP CONDITION: If ANY unchecked, do NOT proceed.
Purpose: Run tests with fix loops to ensure code works correctly.
fix_cycles = 0
MAX_CYCLES = 3
while fix_cycles < MAX_CYCLES:
run test command
if all tests pass:
proceed to Phase 6
break
if tests fail:
analyze failure output
apply targeted fix
fix_cycles += 1
if fix_cycles >= MAX_CYCLES:
escalate to user
After 3 failed fix cycles:
AskUserQuestion(
questions: [{
question: "Tests failed after 3 fix attempts. How do you want to proceed?",
header: "Escalate",
options: [
{label: "Continue trying", description: "Reset counter and try more fixes"},
{label: "Skip tests", description: "Proceed to completion with failing tests"},
{label: "Stop", description: "Pause execution, I'll investigate"}
],
multiSelect: false
}]
)
**Test Phase Gate** (Required):
STOP CONDITION: If ANY unchecked, do NOT proceed.
Purpose: Verify completion and wrap up.
rm -f docs/lighterpowers/current-progress.md
Note: This file is gitignored. Only cleanup AFTER build/test verification.
Display implementation summary:
## One-Shot Implementation Complete
**Tasks Completed:** N/N
**Build Status:** [Pass/Fail]
**Test Status:** [Pass/Fail/Skipped]
**Summary of Changes:**
- [brief summary]
**Discovered Work (for later):**
- [list items if any]
AskUserQuestion(
questions: [{
question: "Implementation complete. How would you like to finish?",
header: "Complete",
options: [
{label: "Commit all", description: "Stage and commit remaining changes"},
{label: "Review changes", description: "Show me a diff before committing"},
{label: "Done", description: "Leave changes as-is, I'll handle it"}
],
multiSelect: false
}]
)
| Violation | Why It's Critical | Recovery |
|---|---|---|
| Pausing mid-implementation for feedback | Defeats one-shot purpose | Continue to completion, pause only at Phase 4 |
| > 3 fix cycles without escalation | Risk of infinite loops | Escalate to user after 3 cycles |
| Skipping build phase entirely | Ships potentially broken code | Run build or get explicit user skip |
| Plain text questions instead of AskUserQuestion | User can't respond via structured UI | Use AskUserQuestion tool |
| Proceeding to tests without user approval | Violates single-checkpoint contract | Wait for Phase 4 response |
| Different tests failing each fix cycle | Indicates architectural issue | Escalate immediately |
No prompt templates needed: One-shot executes directly without subagents.
## Requirements ReminderBefore announcing completion, verify: