Execution Core Standards

Core Purpose

This skill defines universal behavioral protocols for autonomous agent execution.

It answers: "HOW should agents behave?" (not "WHAT should they do").

Skill Contents

Behavioral Protocols:

• references/observation-points.md - Self-verification and evidence collection
• references/auth-gates.md - Authentication error handling
• references/handoff-protocol.md - Standard handoff format
• references/execution-protocol.md - Universal worker execution steps

Core Behavioral Standards

1. Uninterrupted Flow

Agents execute autonomously without pausing for human input.

Protocol: references/observation-points.md

• Agents execute tasks sequentially
• Verify success programmatically via CLI
• Log results in structured format
• Continue to next task without waiting

2. Self-Verification

Every task must be verified automatically.

Standard Pattern:

**Self-Verification Results:**
✓ [Verification 1 passed]
✓ [Verification 2 passed]

Next: Continue to next task

3. Authentication Gates

Authentication errors are NORMAL, not failures.

Protocol: references/auth-gates.md

1. Recognize auth gate (401, 403, "not authenticated")
2. STOP execution (don't retry in loop)
3. Create HANDOFF.md with exact steps
4. EXIT process
5. Resume after human provides credentials

4. Handoff Protocol

Standard format for pausing execution.

Protocol: references/handoff-protocol.md

Use for:

• Authentication gates
• Critical failures
• Ambiguous requirements
• Missing dependencies

Format:

# HANDOFF Required

**Reason**: [AUTH_GATE | CONFLICT | AMBIGUOUS]

**What Happened**: [Description]

**What You Need to Do**: [Specific action]

**Verification**: [How to confirm fix]

**Next Step**: Restart execution after [action]

Usage in Other Skills

Skills reference execution-core for behavioral standards:

Example from project-strategy:

Use execution-core observation-points for self-verification
Use execution-core auth-gates for authentication handling

Example from software-engineering:

Apply execution-core verification protocols during TDD
Use execution-core handoff format for blockers

Key Principles

1. Behavioral Consistency: All builder operations use same execution standards
2. Self-Sufficiency: Agents verify own work programmatically
3. Evidence Collection: All actions logged with verification
4. Non-Blocking: No waiting loops or human checkpoints
5. Clear Handoffs: Standard format when pausing execution

5. Human Interaction Policy (AskUserQuestion)

To ensure Uninterrupted Flow, the following policy applies to human interaction:

• Execution Phase Workers (e.g., Worker Agent): AskUserQuestion is STRICTLY PROHIBITED. Workers must operate autonomously or use the Handoff Protocol if blocked.
• Planning Phase Coordinators (e.g., Director, Orchestrator): AskUserQuestion is ALLOWED for requirements gathering and clarification before execution begins.

This policy ensures that heavy implementation tasks are never interrupted, while initial discovery remains interactive.

Standards Hierarchy

Execution-core sits at the BASE of the behavioral stack:

execution-core (HOW to behave)
    ↓
project-strategy (WHAT documents to create)
software-engineering (WHAT code patterns to use)
    ↓
director/worker agents (Execute using these standards)

Critical: Execution-core is UNIVERSAL. All builder agents must follow these behavioral protocols.