Augmented Git Workflow (High Intensity Flow)
Context
Traditional git workflows (GitFlow, trunk-based, GitHub Flow) assume either single-expertise teams or sequential review gates. AI-augmented development dramatically increases output speed, making synchronization and arbitration the real bottlenecks. A workflow designed for multi-expertise parallel collaboration is needed.
Options considered
Option A - Adapted trunk-based
Trunk-based development with feature flags.
- Advantage: Simple, well-understood
- Disadvantage: Requires CI maturity and feature flag infrastructure from day one; no space for messy expert iteration
Option B - High Intensity Flow (HIF)
A branch model with 3 tiers (main, clusters, expert branches), where clusters represent time-boxed collaborative "workshop" spaces for cross-expertise work.
- Advantage: Handles parallel multi-expertise work, scales from solo to team, explicit arbitration points
- Disadvantage: More complex than trunk-based, experimental/unproven at scale
Decision
HIF (Option B). The workflow provides a single coherent model across the delegation spectrum -- from human teams through mixed human-agent configurations to fully automated pipelines.
Consequences
Positive
- Expert branches allow parallel iteration without blocking each other
- Clusters provide bounded integration scope (time-boxed, named)
- Arbitration points are explicit (cluster promotion to main)
Negative / Risks
- Experimental (v0.1) -- needs field validation
- Added complexity vs simpler workflows for solo developers
Implementation notes
- Full specification:
autograph.pkg.delivery/docs/high-intensity-flow.md - Dry-run validation:
autograph.pkg.delivery/docs/hif-dry-run.md - Agent skill:
augr-git-conventionsimplements HIF branching conventions