task-bootstrap

Zero-Costdiscovery

Mission

Orient the agent, load project context, identify scope and unknowns before any implementation starts.

When to Use

Use when starting a new task with unclear scope, before any implementation begins, when requirements are vague or ambiguous, or when the agent needs to orient itself on what the user actually wants.

Triggers: “start this task”, “I need help with”, “unclear requirements”, “orient yourself”, “what should we do”

Chain-To

After scope is locked, task-bootstrap auto-chains to the highest-confidence downstream instruction:

• system-design
• feature-implement
• evidence-research
• code-review
• strategy-plan
• issue-debug
• code-refactor
• test-verify
• agent-orchestrate
• policy-govern
• enterprise-strategy
• physics-analysis

Required Pre-Conditions

Before scope analysis, prime context through the agent’s existing MCP surface:

• agent-workspace (list / read) for source files in scope
• Serena MCP tools (mcp__serena__list_memories, mcp__serena__get_symbols_overview) — every task-bootstrap response ends with a 🧭 Serena enrichment footer naming the exact call to make. See Serena Integration.

Reactivation Policy

periodic — reactivated periodically in continuous/agent sessions to catch stale context before it causes routing errors.

Example

{
  "request": "I need to add OAuth2 to the admin portal but I'm not sure where to start"
}

Output: Scope document listing known requirements, open unknowns, suggested next instruction (system-design or feature-implement), and context gaps to resolve.