apex

**Domain**: Execution

Capture learnings (problems, decisions, gotchas) from sessions to make future agents more effective

Systematic debugging with APEX pattern learning

Run full APEX workflow (research -> plan -> implement -> ship)

Run APEX implement phase - build and validate code

Run APEX plan phase - transform research into architecture

Run APEX research phase - gather intelligence via parallel agents

Run iterative adversarial review on a task plan before implementation

Comprehensive adversarial code review using multi-agent system

Run APEX ship phase - review, commit, and reflect

Searches project markdown documentation for relevant context, past decisions, and historical learnings

Predicts likely failures from historical patterns (failures.jsonl). Called by intelligence-gatherer for risk analysis.

**Agent Type**: orchestrator

Mines git history for structured intelligence (commits, regressions, ownership trends). Called by orchestrators during research and execution phases.

Extracts concrete implementation patterns from the current codebase with file:line references and reusable code examples

Queries APEX MCP tools for patterns and task context, loads relevant code files, and synthesizes execution strategy. Focused on APEX database intelligence - git/risk/systems analysis handled by dedicated agents.

Searches past task files for relevant learnings, problems solved, decisions made, and gotchas discovered

Mines codebase to discover new reusable patterns (3+ occurrences). Use after successful implementations to build pattern library.

Performs multi-lens code review (correctness, maintainability, resilience, patterns). Use before PR or at milestones.

A production-grade code review system that uses adversarial agents to eliminate false positives while maintaining thoroughness.

Validate architectural integrity, design patterns, and system consistency in code changes

Assess code readability, maintainability, complexity, and adherence to coding standards

Analyze git history for pattern violations, regressions, and codebase inconsistencies

Identify security vulnerabilities in code changes with evidence-based analysis

Assess test coverage quality and identify untested code paths and edge cases

Unified adversarial challenger - validates findings, checks history, assesses ROI, and can override scores

Surfaces novel risks and edge cases using structured reasoning. Complements historical failure data with forward-looking analysis.

Analyzes complex systems to map dependencies, trace execution flows, and explain architectural relationships.

Executes comprehensive testing with predictive analysis. Runs syntax checks, linting, tests, and identifies failure patterns.

Conducts web research with source verification. Use for external knowledge, API docs, current events, or fact validation.

Capture learnings (problems, decisions, gotchas) from sessions to make future agents more effective

Systematic debugging with pattern learning. Applies hypothesis-driven investigation, evidence collection, and reflection to update pattern confidence.

Orchestrator that runs the full APEX workflow (research → plan → implement → ship) in a single session. Use for tasks you want to complete without context switches.

Build and validate loop (BUILDER + validation) - implements the architecture, runs tests, iterates until passing. Writes code following the plan.

Architecture phase (ARCHITECT) - transforms research into rigorous technical architecture through 5 mandatory design artifacts. Interactive and iterative.

Intelligence gathering phase - spawns parallel agents to analyze codebase, patterns, git history, and web research. Creates or updates task file with findings.

Use when a task already has research and plan sections and the plan needs adversarial review before implementation

Review and finalize (REVIEWER + DOCUMENTER phases) - runs adversarial code review, commits changes, completes task, and records reflection to capture pattern outcomes.