Now liveThe Skillselion MCP - thousands of ranked skills, loaded into your agent mid-task. No install.Get it →
othmanadi avatar

Apex Tier2

  • 4 installs
  • 2 repo stars
  • Updated March 14, 2026
  • othmanadi/apex

Execute a complex multi-file task with the agent working autonomously between human checkpoints for steering and review.

About

Executes a complex, multi-file task autonomously but pauses at predefined checkpoints for human steering and review. A developer uses it for medium-complexity feature work or refactoring that needs judgment calls along the way.

  • Plans phases with human checkpoints between data, logic, API, and tests
  • Works autonomously between checkpoints, pausing for steering

Apex Tier2 by the numbers

  • 4 all-time installs (skills.sh)
  • +1 installs in the week ending Aug 4, 2026 (Skillselion tracking)
  • Ranked #1,780 of 2,715 Automation & Workflows skills by installs in the Skillselion catalog
  • Data as of Aug 4, 2026 (Skillselion catalog sync)
npx skills add https://github.com/othmanadi/apex --skill apex-tier2

Add your badge

Show developers this skill is listed on Skillselion. Paste this into your README.

Listed on Skillselion
Installs4
repo stars2
Last updatedMarch 14, 2026
Repositoryothmanadi/apex

What it does

Execute a complex multi-file task with the agent working autonomously between human checkpoints for steering and review.

Files

SKILL.mdMarkdownGitHub ↗

APEX Tier 2 — Guided Execution

Work autonomously on a complex task but pause at predefined checkpoints for human steering. The agent does the heavy lifting; the human makes the judgment calls.

When to Use

Use Tier 2 when:

  • The task involves multiple files or modules
  • Some decisions require human judgment (naming, architecture choices)
  • The work takes more than one focused session
  • You want to review progress before the agent goes further

Escalate to Tier 3 when:

  • Every step requires discussion
  • The architecture is being designed from scratch
  • The task is exploratory with no clear end state

Workflow

1. Plan the work and define checkpoints 2. Execute until the next checkpoint 3. Present checkpoint summary for human review 4. Incorporate feedback and continue 5. Final validation and PR

Step 1: Plan and Define Checkpoints

Break the task into phases. Each phase ends with a checkpoint:

## Execution Plan

### Phase 1: Data layer changes
- Modify schema for new fields
- Update migrations
- **CHECKPOINT: Review schema changes before proceeding**

### Phase 2: Business logic
- Implement service methods
- Add validation rules
- **CHECKPOINT: Review logic before wiring to API**

### Phase 3: API layer
- Add/modify endpoints
- Update request/response types
- **CHECKPOINT: Review API contract before tests**

### Phase 4: Tests and cleanup
- Write unit and integration tests
- Update documentation
- Final validation

Present this plan to the user. Get approval before starting.

Step 2: Execute Until Checkpoint

Work through the current phase. Follow the same implementation rules as Tier 1: one file at a time, match specs, preserve edge cases.

Run the self-correcting loop after each phase:

Linux/Mac:

bash "${CLAUDE_SKILL_DIR}/scripts/validate.sh" <target-dir>

Windows:

powershell -File "${CLAUDE_SKILL_DIR}/scripts/validate.ps1" <target-dir>

Step 3: Checkpoint Summary

When reaching a checkpoint, present a structured summary:

## Checkpoint: {Phase Name}

### Completed
- [x] Item 1
- [x] Item 2

### Decisions Made
| Decision | Choice | Reasoning |
|----------|--------|-----------|
| Field naming | `created_at` | Matches existing convention |

### Questions for You
1. Should X use pattern A or pattern B?
2. Is the error message "..." appropriate for this context?

### Next Phase Preview
Phase 2 will implement [brief description]. Estimated scope: N files.

Wait for human response before continuing. Do not proceed past a checkpoint without explicit approval.

Step 4: Incorporate Feedback

Apply the human's feedback. If feedback contradicts a previous decision, update the relevant code AND update AGENTS.md if the feedback reveals a recurring preference (two-strike rule).

Step 5: Final Validation and PR

After all phases complete, run full validation and open a PR with the complete execution log:

gh pr create --draft --title "apex: {task description}" --body-file /tmp/apex-pr-body.md

Output

A draft PR with a complete audit trail of checkpoints, decisions, and human feedback incorporated at each stage.

Example PR body with checkpoint log:

## apex: Refactor payment module to Stripe v3

### Execution Log

#### Phase 1: Data Layer ✅
- Updated PaymentIntent schema
- Added new webhook event types
- **Human feedback:** "Use camelCase for new fields" → Applied

#### Phase 2: Business Logic ✅
- Migrated to PaymentIntents API
- Added idempotency keys
- **Human feedback:** "Add retry logic for network errors" → Applied

#### Phase 3: API Layer ✅
- Updated `/payments/create` endpoint
- Added `/payments/confirm` endpoint
- **Human feedback:** None needed

#### Phase 4: Tests ✅
- 23 new tests, 0 failing
- Updated 8 existing tests

### Validation: ✅ All checks pass

Related skills

Automation & Workflowsagentsautomation

This week in AI coding

Five minutes, every Monday - the tools, releases and tactics for developers.

unsubscribe anytime.