apex-tier3
APEX Tier 3 — Pair Mode
Work side-by-side with the human in real-time. Every significant decision is discussed. The agent proposes, the human approves or redirects.
When to Use
Use Tier 3 when:
- Designing new architecture from scratch
- Debugging a critical production issue
- The problem space is ambiguous or exploratory
- The human wants to learn from the process
- Stakes are high and mistakes are expensive
Interaction Protocol
The Propose-Confirm Loop
Every action follows this pattern:
More from othmanadi/apex
apex-learn
Trigger the self-improvement loop after completing a task. Analyzes friction, extracts learnings, and updates AGENTS.md with new rules and preferences. Use after any task that involved struggle, correction, or discovery of a new pattern. Triggers on keywords like learn, self-improve, update rules, what did I learn, retrospective, post-mortem.
2apex-replatform
Implement features in a new architecture using specifications from apex-decompose. Use when replatforming, migrating codebases, rebuilding features in a new framework, or translating specs into working code. Triggers on keywords like replatform, migrate, rebuild, implement spec, new architecture.
2apex-tier1
Execute well-defined tasks autonomously with zero human intervention. Fire-and-forget mode for straightforward work like adding tests, fixing lint errors, writing docs, or implementing simple features. Use when the task is clear, low-risk, and has objective pass/fail criteria. Triggers on keywords like fire and forget, autonomous task, simple fix, add tests, fix lint.
2apex-decompose
Decompose features from an existing codebase into detailed, agent-ready specifications for replatforming or migration. Use when analyzing legacy code, extracting feature behavior, mapping old architecture to new, or preparing migration specs. Triggers on keywords like decompose, analyze features, extract behavior, migration spec, map codebase.
2apex-tier2
Execute complex tasks with periodic human checkpoints. Background work with steering — the agent works autonomously but pauses at defined decision points for human input. Use for medium-complexity tasks like feature implementation, refactoring, or multi-file changes. Triggers on keywords like guided task, checkpoint, background with review, implement feature, refactor module.
2