.claude/skills/execute/SKILL.md
Execute a plan. Creates ./.gtd/<task_name>/{phase}/SUMMARY.md
npx skillsauth add Hoang604/get-thing-done executeInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
Core responsibilities:
Flow: Load Plan → Execute Tasks (Apply Code Standards) → Verify → Summarize </objective>
<context> **Phase number:** $ARGUMENTSRequired files:
./.gtd/<task_name>/{phase}/PLAN.md — Must existOutput:
./.gtd/<task_name>/{phase}/SUMMARY.md<standards_and_constraints>
<execution_philosophy>
Execute one task fully before moving to the next.
After each task, check its done criteria. Don't proceed if verification fails.
Implement exactly what the plan specifies. No more, no less. If you think the plan is wrong:
<code_principles> Mantra: "Code is not an asset; it is a liability. Every line must earn its place."
| Zone | Trust Level | Action | | ------------------------------ | ----------- | --------------------- | | Edge (API, user input, DB) | ZERO trust | Validate everything | | Core (internal logic) | HIGH trust | Skip redundant checks |
Empty catch blocks are forbidden.
Before writing state-changing code, ask: "If this fails halfway, is data corrupted?"
finally for cleanupEvery number, string, or value must have a name.
</code_principles>
<deviation_policy> | Situation | Action | | -------------------------- | ------------------------ | | Small bug found | Auto-fix | | Missing dependency | Install, note in summary | | Unclear requirement | STOP, ask user | | Architecture change needed | STOP, ask user | </deviation_policy>
<prohibitions> - **NEVER** deviate from plan silently - **NEVER** swallow errors (no empty catch blocks) - **NEVER** use `any` type (unless absolutely unavoidable) - **NEVER** implement without reading dependencies first - **NEVER** scatter retry logic </prohibitions></standards_and_constraints>
<process>Bash:
if ! test -f "./.gtd/<task_name>/$PHASE/PLAN.md"; then
echo "Error: No plan exists for phase $PHASE"
exit 1
fi
Read ./.gtd/<task_name>/$PHASE/PLAN.md.
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
GTD ► EXECUTING PHASE {N}
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Objective: {objective}
Tasks:
[ ] 1. {task 1 name}
[ ] 2. {task 2 name}
─────────────────────────────────────────────────────
Loop through each task in PLAN.md:
► Task {N}: {name}
Files: {files}
Before calling any existing function/library:
Implement the task using <code_principles>.
any).Check the task's <done> criteria.
If verified:
✓ Task {N} complete
If not verified:
Note any work done outside the plan (bugs fixed, adjustments made) for the Summary.
After all tasks, check plan's success criteria:
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
GTD ► VERIFYING PHASE {N}
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
[✓] {criterion 1}
[✓] {criterion 2}
If any fail: Attempt to fix or ask user.
Write to ./.gtd/<task_name>/$PHASE/SUMMARY.md:
# Phase {N} Summary
**Status:** Complete
**Executed:** {date}
## What Was Done
{Narrative summary of implementation}
## Behaviour
**Before:** {describe system behaviour before changes}
**After:** {describe system behaviour after changes}
## Tasks Completed
1. ✓ {task 1 name}
- {what was implemented}
- Files: {files changed}
2. ✓ {task 2 name}
...
## Deviations
{List any work done outside the plan, or "None"}
## Success Criteria
- [x] {criterion 1}
- [x] {criterion 2}
## Files Changed
- `{file 1}` — {what changed}
## Proposed Commit Message
feat(phase-{N}): {short description}
{longer description if needed}
- {bullet 1}
- {bullet 2}
Update ./.gtd/<task_name>/ROADMAP.md phase status:
### Phase {N}: {Name}
**Status**: ✅ Complete
</process>
<offer_next>
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
GTD ► PHASE {N} COMPLETE ✓
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Summary written to: ./.gtd/{N}/SUMMARY.md
Tasks: {X}/{X} complete
Deviations: {count}
Files changed: {count}
─────────────────────────────────────────────────────
▶ Next Up
/plan {N+1} — plan the next phase
─────────────────────────────────────────────────────
</offer_next>
<forced_stop> STOP. The workflow is complete. Do NOT automatically run the next command. Wait for the user. </forced_stop>
testing
manual trigger by user, do not auto invoke
tools
manual trigger by user, do not auto invoke
development
Trace execution paths and document how code actually behaves. Use when you need to understand how features work, walk through code flows, explain component behavior, trace where data comes from, understand relationships between components, or audit for orphaned events and dead code.
testing
Guide users through a structured workflow for co-authoring documentation. Use when user wants to write documentation, proposals, technical specs, decision docs, or similar structured content. This workflow helps users efficiently transfer context, refine content through iteration, and verify the doc works for readers. Trigger when user mentions writing docs, creating proposals, drafting specs, or similar documentation tasks.