iterative-retrieval

Refine codebase context for subagents through a four-phase retrieval loop.

Updated Mar 1, 2026

Search Tags:#codebase #multi-agent #dispatch #context-management #evaluate #iterative-retrieval #retrieval-pipeline

Ready to run

npx skills add https://github.com/derekhu0002/ai4pb-orchestrator --skill iterative-retrieval-derekhu0002

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: iterative-retrieval
Source: https://github.com/derekhu0002/ai4pb-orchestrator/tree/main/skills/iterative-retrieval
Command: npx skills add https://github.com/derekhu0002/ai4pb-orchestrator --skill iterative-retrieval-derekhu0002

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Addresses the context provisioning challenge in multi-agent workflows by progressively refining the contextual data exposed to subagents, reducing token bloat while keeping essential signals.

Core Features & Use Cases

Progressive context refinement pipeline for subagents operating on large codebases.
Automatic surface of high-relevance files and explicit identification of remaining context gaps.
Safe, bounded loop (max three cycles) with clear stop criteria for reliable agent orchestration.
Suitable for tasks like code exploration, reverse engineering, or dynamic task delegation where context evolves over time.

Quick Start

Start by performing a broad context search, then iteratively evaluate and refine results, looping up to three cycles to assemble a high-quality context set for subagents.

Frequently Asked Questions about iterative-retrieval

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How does progressive context refinement work in multi-agent workflows?▼

Progressive context refinement in multi-agent workflows uses a four-phase loop: Dispatch, Evaluate, Refine, and Loop. This mechanism progressively surfaces high-relevance files and identifies remaining knowledge gaps for subagents.

What is the best way to stop context bloat when spawning subagents with limited initial context?▼

To stop context bloat when spawning subagents with limited initial context, apply a bounded refinement loop. This approach safely limits iteration to a maximum of three cycles with clear stop criteria.

Can I use iterative retrieval for dynamic task delegation where context evolves over time?▼

Yes, iterative retrieval supports dynamic task delegation where context evolves over time. It progressively refines codebase context, making it suitable for code exploration and reverse engineering tasks.

What are the limitations of using a bounded loop for codebase context management?▼

The main limitation of bounded loop codebase context management is the strict maximum cycle constraint. The loop stops after three cycles, which may leave remaining knowledge gaps if the codebase is exceptionally complex.