Autonomous Coding Workflow
SkillDev toolsLets your agent make code changes across a repository in small verified steps with tests and checks after each edit.
Available today. Use it from your connected AI after setup.
No other account needed.
Connect ahel once, and every AI you use reads what you have installed.
Then ask your AI: use the Autonomous Coding Workflow skill
About this capability
Governs safe, incremental, repository-wide development workflow with continuous validation gates
What this skill tells your AI
The instructions your AI receives, as published by bottelet/daybydaycrm in .claude/skills/autonomous-coding-workflow/SKILL.md and read by ahel’s review.
Goal
Perform repository-wide modifications safely, incrementally, and with continuous validation.
1. Instruction Precedence
Before doing anything:
- Check for repository-level instruction files:
.github/copilot-instructions.mdAGENTS.md.junie/*.mdCLAUDE.md
If they exist:
- Treat them as higher precedence for architecture and conventions.
- Avoid duplicating rules already defined there.
2. Preparation
Before modifying code:
- Read existing implementation.
- Understand current behavior.
- Identify existing abstractions and reuse them:
- Traits
- Base test cases
- Base resources
- Base seeders
- Services
- DTOs
- Transformers
- Search explicitly for duplication before introducing new abstractions.
- Preserve existing architectural patterns.
Do not modify code that has not been understood.
3. Refactoring Heuristics
Apply only when relevant:
- If repeated patterns exist across many test classes, models, or resources, evaluate abstraction opportunities.
- Prefer centralizing duplicated logic into:
- Traits
- Base classes
- Services
- Do not introduce abstraction unless duplication is confirmed.
4. Incremental Development
Work in small, verifiable steps.
After each change:
- Verify syntax:
php -l - Run targeted tests.
- Fix failures immediately.
- Run code style checks:
vendor/bin/pint --dirty --format agent - Continue only if repository is clean.
5. Validation Gates
Never proceed if any of the following fail:
- PHPUnit tests
- Static analysis
- PHP syntax check (php -l)
- Code style (Pint)
Before completion, additionally ensure:
- migrate:fresh --seed passes
- smoke tests pass
- targeted tests pass
- full suite passes (unless explicitly excluded)
6. Module Completion
After completing a module:
- Run targeted test suite.
- Run
php -l. - Run Pint.
- Confirm no unintended changes.
- Commit with clear module description.
Do not start the next module until the current one is fully stable.
7. Uncertainty Handling
If behavior is unclear:
- Stop immediately.
- Describe ambiguity.
- Request clarification.
- Do not infer or guess missing business rules.
8. Success Criteria
The task is complete only when:
- Behavior is preserved.
- No duplicate logic introduced.
- Changes are idempotent.
- All tests pass.
- Full suite passes.
- Formatting is clean.
- No unintended architectural drift occurred.
Signals
- GitHub stars
- 2k
- Forks
- 767
- Last commit
- Aug 2026
Advanced
- Catalog kind
- skill
- Gateway key
autonomous-coding-workflow- Source
- github.com/bottelet/daybydaycrm