dispatching-parallel-agents

name	dispatching-parallel-agents
description	Use when deciding whether to split work across parallel agents or keep it sequential for audits, investigations, implementation slices, reviews, verification, repository analysis, debugging, or multi-domain APIVR work. Provides dispatch decision rules, model-selection tiers, prompt fields, and safe parallelization guardrails.
activation	Activate when the description trigger applies to the current task.
required_inputs	Task request, relevant repository context, constraints, and authority dependencies.
required_outputs	Skill-specific artifact, verification evidence, canonical verdict, and next action.
authority_dependencies	00_start_here/SOURCE_OF_TRUTH.md; 10_governance/APIVR_EXECUTION_LIFECYCLE.md; 10_governance/source_of_truth/Elite_Build_Goals_v3.md.
evidence_requirements	Executed checks or an honest Unknown, Not Run, or Blocked state for every material claim.

Dispatching Parallel Agents

Use this skill before launching subagents or parallel workstreams.

Parallel dispatch must reduce risk, time, or context load without fragmenting source-of-truth control. APIVR remains centralized with one evidence ledger and one final verdict.

Dispatch Decision Graph

flowchart TD
  A["Task begins"] --> B{"Can one agent safely finish within context and time?"}
  B -- "Yes" --> C["Stay sequential"]
  B -- "No" --> D{"Are work slices independent or read-only?"}
  D -- "No" --> E["Stay sequential or create strict handoff order"]
  D -- "Yes" --> F{"Is shared state protected by exact scopes?"}
  F -- "No" --> G["Define scopes before dispatch"]
  F -- "Yes" --> H{"Does risk justify specialist review?"}
  H -- "No" --> I["Dispatch cheap/standard research or verification agents"]
  H -- "Yes" --> J["Dispatch capable specialist agents plus review gate"]
  I --> K["Merge evidence through APIVR"]
  J --> K

When To Dispatch

Dispatch when at least one condition is true:

• Multiple independent files, modules, providers, or workflows must be audited.
• One agent needs implementation while another performs independent review.
• A forensic or comprehensive audit needs specialist perspectives.
• Context would become overloaded if all evidence stayed in one agent.
• Time matters and work can be safely split without write conflicts.

Stay sequential when:

• The task is small and local.
• Correctness depends on one continuous reasoning chain.
• Work slices would modify the same files without a strict owner.
• Security, data, or release authority is unclear.

Model Tiers

Tier	Use for	Avoid for
Cheap	File inventory, link checks, simple search, formatting review, duplicate detection.	Architecture, security, data loss, payment, auth, final verdicts.
Standard	Focused implementation, unit tests, local refactors, component review, docs wiring.	Forensic judgment, ambiguous design, high-risk release calls.
Capable	Architecture, security, data integrity, production incidents, final review, synthesis.	Bulk mechanical scans that cheaper agents can do.

Dispatch Prompt Fields

Every dispatched prompt must include:

Role:
APIVR tier:
Objective:
Exact scope:
Files/systems allowed:
Files/systems forbidden:
Source-of-truth files:
Applicable Elite Build Goals:
Required skill(s):
Evidence required:
Stop conditions:
Output format:
Status allowed: DONE / DONE_WITH_CONCERNS / NEEDS_CONTEXT / BLOCKED