zte-progression
Guide progression from In-Loop to Out-Loop to Zero-Touch Engineering. Use when assessing agentic maturity, planning ZTE progression, or identifying requirements for autonomous operation.
When & Why to Use This Skill
The ZTE Progression skill is a strategic framework designed to guide engineering teams through the evolution of agentic workflows. It facilitates the transition from manual 'In-Loop' prompting to 'Zero-Touch Engineering' (ZTE) by providing structured assessments of agentic maturity, identifying technical blockers, and establishing clear KPIs for autonomous operation. This skill is essential for organizations looking to scale AI agent autonomy safely and effectively within their software development lifecycle.
Use Cases
- Maturity Assessment: Evaluating whether current engineering workflows are In-Loop, Out-Loop, or ready for Zero-Touch automation based on presence KPIs and human intervention frequency.
- Strategic Roadmap Planning: Developing a step-by-step progression plan to move from manual reviews to fully autonomous 'shipping' for specific task classes like chores or bug fixes.
- Blocker Identification: Analyzing system weaknesses such as low test coverage, inconsistent success rates, or lack of rollback capabilities that prevent the adoption of ZTE.
- Confidence Building & KPI Tracking: Monitoring success streaks and review catch rates to provide data-driven justification for increasing agent autonomy.
| name | zte-progression |
|---|---|
| description | Guide progression from In-Loop to Out-Loop to Zero-Touch Engineering. Use when assessing agentic maturity, planning ZTE progression, or identifying requirements for autonomous operation. |
| allowed-tools | Read, Grep, Glob |
ZTE Progression Skill
Guide teams through the three levels of agentic coding: In-Loop, Out-Loop, and Zero-Touch Engineering.
When to Use
- Assessing current agentic maturity level
- Planning progression to higher automation
- Identifying blockers to ZTE adoption
- Building confidence for autonomous operation
Core Concepts
Three Levels
| Level | Presence KPI | Description |
|---|---|---|
| In-Loop | Constant | Interactive prompting |
| Out-Loop | 2 (prompt + review) | AFK agents with review |
| Zero-Touch | 1 (prompt only) | Full automation |
Progression Requirements
In-Loop -> Out-Loop:
- Workflows succeed on first attempt
- Tests catch issues reliably
- Similar tasks repeat frequently
Out-Loop -> ZTE:
- 90%+ success rate
- Review catches nothing new
- Tests provide safety net
Assessment Workflow
Step 1: Identify Current Level
Search for agentic workflow indicators:
Look for:
- Manual prompting patterns (In-Loop)
- Automated workflows with review (Out-Loop)
- Fully automated shipping (ZTE)
Step 2: Analyze Success Metrics
Review KPI data if available:
Check @agentic-kpis.md or equivalent:
- Attempt counts
- Success streaks
- Review catch rates
Step 3: Identify Blockers
Common blockers to progression:
| Blocker | Mitigation |
|---|---|
| Low test coverage | Improve tests before progressing |
| Inconsistent success | Analyze failures, improve prompts |
| Review catches issues | Tests need to catch these first |
| No rollback capability | Add before enabling ZTE |
Step 4: Recommend Next Steps
Based on current level, recommend:
At In-Loop:
- Start with Out-Loop for chores
- Build workflow automation
- Establish review process
At Out-Loop:
- Track review catch rate
- Build confidence with simple tasks
- Consider ZTE for high-confidence areas
Near ZTE:
- Enable for chores first
- Expand progressively
- Monitor continuously
Key Memory References
- @zte-progression.md - Three levels definition
- @zte-confidence-building.md - Building ZTE confidence
- @agentic-kpis.md - KPI tracking
- @composable-primitives.md - Workflow building blocks
Output Format
Provide assessment:
## ZTE Assessment
**Current Level:** [In-Loop | Out-Loop | ZTE]
**Problem Classes Assessed:** [chores, bugs, features]
### Indicators Found
- [Evidence of current level]
### Blockers to Progression
- [Identified blockers]
### Recommendations
1. [Specific next step]
2. [Specific next step]
### Target Timeline
- [Realistic progression timeline]
Anti-Patterns to Identify
- Skipping levels (Out-Loop before tests exist)
- All-or-nothing thinking (must ZTE everything)
- Ignoring failure signals (pushing forward despite issues)
- Review theater (reviewing without catching anything)
Version History
- v1.0.0 (2025-12-26): Initial release
Last Updated
Date: 2025-12-26 Model: claude-opus-4-5-20251101