| name | autonomous-dev |
| description | Harness for autonomous software development. Use when fixing or remediating known issues (including security vulnerabilities). Enforces lifecycle through alignment gates (PROJECT.md), adversarial generator/evaluator agents, and autonomous orchestration of project issues. NOT for setting up standalone CI/CD pipelines. |
Autonomous Development Workflow
This skill provides a deterministic software engineering harness, designed to wrap AI code generation with strict alignment gates, adversarial evaluations, CI/CD pipeline principles, and collaborative Git practices. Load this skill when designing or orchestrating multi-step architectural features, automating deployment workflows, or writing code inside a structured environment.
1. Project Alignment Gate (PROJECT.md)
Before beginning implementation, you MUST cross-reference the proposed feature against PROJECT.md at the repository root. If PROJECT.md is missing, ask the user to clarify the project goals or constraints before proceeding.
- In-Scope: Confirm the feature directly serves a goal in
PROJECT.md. Proceed.
- Out-of-Scope: Hard block. Halt work immediately and inform the user.
- Constraints: Abide strictly by the constraints (language, architecture, dependencies) defined in the project file.
2. Security Safety Nets
- Execution Sandboxing: Never run unknown or unverified shell scripts directly. Execute isolated tasks within the
tmp/ scratch directory.
- Secret Protection: Check for exposed credentials or PII in diffs before committing. Never commit
.env.
- Command Gates: Do not execute destructive commands (
rm -rf, git reset --hard) without explicit Human-in-the-Loop approval.
3. Generator / Evaluator Pipeline
Features cannot simply be written and committed. They must navigate a rigid pipeline based on the adversarial evaluation pattern:
- Research: Analyze local scope and external dependencies via parallel search strategies.
- Planning: Formulate architecture and acceptance criteria (generator).
- Acceptance Tests FIRST: Tests must be written before main implementation (TDD).
- Implementation: Produce code.
- HARD GATE: 0 test failures. You cannot proceed if tests fail. No stubs/placeholders allowed.
- Adversarial Review: Self-evaluate the code as a skeptical reviewer (evaluator). Actively search for edge cases, security vulnerabilities, efficiency loss, and anti-patterns.
- Documentation: Ensure docs stay tightly in sync with the codebase after the feature clears CI.
4. Workflow Orchestration
Follow strict version control (Conventional Commits, atomic commits) and CI/CD automation rules. Test code locally before creating PRs.
5. Context Management & Drift Prevention
As the context window fills up, context anxiety can degrade performance.
- Recognize when the session has spanned too many features or files (e.g., beyond 4-5 features).
- Use session summaries or persistent memories to bookmark state, clear the context, and resume with a fresh perspective.
6. Output Templates
Generator Plan Output:
### Feature Plan: [Feature Name]
**Scope:** [In-Scope/Out-of-Scope based on PROJECT.md]
**Constraints:** [Key constraints to respect]
**Acceptance Criteria:**
- [ ] Criterion 1
- [ ] Criterion 2
Evaluator Review Output:
### Adversarial Review: [Feature Name]
**Pass/Fail:** [Result]
**Security Risks:** [Identified risks or "None"]
**Edge Cases Missed:** [Edge cases]
**Required Fixes:** [List of fixes before proceeding]
Synergies
| Skill/Workflow | Relationship |
|---|
github-commander | Complementary — handles issue orchestration while this handles the dev lifecycle |
adversarial-planner | Can be used during the Planning phase of the Generator pipeline |
adversarial-security | Can be used during the Evaluator phase |