How do I install the Openspec Apply skill?

Run npx skillmds@latest add freddy-chen-ux/openspec-apply 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 Openspec Apply safe to use?

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

Which AI agents work with Openspec Apply?

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 Openspec Apply free to use?

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

Who published Openspec Apply?

Freddy-CHEN-ux (@freddy-chen-ux) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Openspec Apply by freddy-chen-ux — Agent Skill

Name: Openspec Apply
Rating: 4.666666666666667 (3 reviews)
Author: Freddy-CHEN-ux

File contents

OpenSpec Apply

Implement approved changes by following the task list exactly.

Announce: "I'm using openspec-apply to implement change [id] following the approved tasks."

Iron Law

FOLLOW THE TASK LIST EXACTLY - NO ADDITIONS, NO SKIPS

The tasks were approved. Your job is to execute them, not improve them.

Process

Phase 1: Load Context

Read the proposal:
```
openspec show [change-id]
```
Read tasks.md completely before starting
Identify current progress:
- Which tasks are [x] complete?
- Which is the next [ ] pending task?

Phase 2: Execute Tasks

For EACH task:

Step 1: Announce

Starting Task N.M: [task description]

Step 2: Load TDD skill → REQUIRED SUB-SKILL: Load test-tdd for any code task

Step 3: Follow task steps exactly

Create/modify the exact files listed
Write the exact code specified
Run the exact verification commands

Step 4: Verify

Test passes?
No other tests broken?
Types check?

Step 5: Mark complete Update tasks.md: - [ ] → - [x]

Step 6: Commit

git add [files]
git commit -m "[type](scope): [description]"

Phase 3: Report Progress

After each task:

Completed Task N.M: [description]
- Files changed: [list]
- Tests: PASS
- Next: Task N.M+1

Phase 4: Completion

When all tasks are [x]:

Run full verification:

bun run type-check
bun run lint
bun test
supabase test db

Report:

All tasks complete for change [id].
- [X] tasks completed
- All tests passing
- Ready for code review

Load code-review skill for final review

Working with Subagents

If dispatching domain experts:

Load subagent-workflow skill
Give them the specific task from tasks.md
Track their session_id
Verify their output
If issues found, recall their session for fixes

REQUIRED SUB-SKILL

For any code task → Load test-tdd
For subagent orchestration → Load subagent-workflow
After completion → Load code-review

Red Flags - STOP

If you catch yourself:

Adding features not in tasks
Skipping tasks "to save time"
Implementing differently than specified
Not running verification steps
Marking tasks complete without verification

STOP. Follow the task list.

Common Rationalizations

Excuse	Reality
"I see a better way"	The plan was approved. Follow it.
"This task is unnecessary"	The plan was approved. Do it anyway.
"I'll combine these tasks"	Tasks are sized deliberately. Do them separately.
"Tests are passing, skip verification"	Run ALL verification. Every time.

What To Do When Stuck

If a task is unclear or blocked:

STOP - do not guess
State what's unclear
Ask for clarification
Wait for response

Do NOT proceed with assumptions.

---name: openspec-applydescription: Use when implementing an APPROVED OpenSpec change. Read tasks.md and follow EXACTLY. Do not deviate, add features, or skip tasks. Use with test-tdd for each code task.---# OpenSpec ApplyImplement approved changes by following the task list exactly.> **Announce:** "I'm using openspec-apply to implement change [id] following the approved tasks."## Iron Law```FOLLOW THE TASK LIST EXACTLY - NO ADDITIONS, NO SKIPS```The tasks were approved. Your job is to execute them, not improve them.## Process### Phase 1: Load Context1. Read the proposal: ```bash openspec show [change-id] ```2. Read `tasks.md` completely before starting3. Identify current progress: - Which tasks are `[x]` complete? - Which is the next `[ ]` pending task?### Phase 2: Execute TasksFor EACH task:**Step 1: Announce**```Starting Task N.M: [task description]```**Step 2: Load TDD skill**→ REQUIRED SUB-SKILL: Load `test-tdd` for any code task**Step 3: Follow task steps exactly**- Create/modify the exact files listed- Write the exact code specified- Run the exact verification commands**Step 4: Verify**- Test passes?- No other tests broken?- Types check?**Step 5: Mark complete**Update `tasks.md`: `- [ ]` → `- [x]`**Step 6: Commit**```bashgit add [files]git commit -m "[type](scope): [description]"```### Phase 3: Report ProgressAfter each task:```Completed Task N.M: [description]- Files changed: [list]- Tests: PASS- Next: Task N.M+1```### Phase 4: CompletionWhen all tasks are `[x]`:1. Run full verification: ```bash bun run type-check bun run lint bun test supabase test db ```2. Report: ``` All tasks complete for change [id]. - [X] tasks completed - All tests passing - Ready for code review ```3. Load `code-review` skill for final review## Working with SubagentsIf dispatching domain experts:1. Load `subagent-workflow` skill2. Give them the specific task from tasks.md3. Track their session_id4. Verify their output5. If issues found, recall their session for fixes## REQUIRED SUB-SKILL- For any code task → Load `test-tdd`- For subagent orchestration → Load `subagent-workflow`- After completion → Load `code-review`## Red Flags - STOPIf you catch yourself:- Adding features not in tasks- Skipping tasks "to save time"- Implementing differently than specified- Not running verification steps- Marking tasks complete without verificationSTOP. Follow the task list.## Common Rationalizations| Excuse | Reality ||--------|---------|| "I see a better way" | The plan was approved. Follow it. || "This task is unnecessary" | The plan was approved. Do it anyway. || "I'll combine these tasks" | Tasks are sized deliberately. Do them separately. || "Tests are passing, skip verification" | Run ALL verification. Every time. |## What To Do When StuckIf a task is unclear or blocked:1. STOP - do not guess2. State what's unclear3. Ask for clarification4. Wait for responseDo NOT proceed with assumptions.

Frequently asked questions

Openspec Apply