Skip to content
Directory
GitHub Star

Module Deep Dives ​

Follow module objects, interfaces, lifecycles and boundaries across six layers. Only module-focused essays belong in L1–L6; cross-module topics and engineering retrospectives remain related reading and do not count as module articles.

Local implementation and offline checks do not establish production readiness, real-model performance or financial research quality. Each article retains its specific limits and outstanding checks. Links marked “Chinese original” lead to articles without a published English edition.

IN THIS CATEGORY41 articles

Counts include articles in this category only; related reading is listed separately.

Tags · match all selected
01 / 06

L1 · Task Input and Triggers

Scheduled tasks and other authorized triggers start work; input does not replace research decisions.

  1. Scheduled Research Beyond a Conversation
  2. Daily Personalized Research Is Not Just Repeating a Prompt
02 / 06

L2 · Identity, Sessions and Context

Identity, capability admission, context and durable interactions follow their primary Owners across layers.

  1. Keep Chatting Is Not Keep Executing: Agent Conversation Admission and Task Continuation
  2. How Does an Agent Keep Each User’s Research Private?
  3. Models Understand; Code Authorizes: Run-Scoped Authorization Through Recovery
  4. Extension Host: From Feature Phone to Smartphone
  5. Main Context Compaction Without Rewriting Facts
  6. Cross-Run Compaction: Why an Agent Cannot Simply Summarize History
  7. Why Agents Need a Workspace: From One-Off Conversations to Long-Term Research
  8. Memory: From Saving and Recall to Safe Forgetting
  9. Prompt Injection Is More Than Malicious Text: An Agent Content Trust Boundary
  10. Why an Agent Cannot Execute a Stop Command Immediately
  11. After an Agent Asks: Durable Human Response Adoption
  12. A Custom Expert Should Be More Than a Prompt Field
  13. Agent Usage: Requests, Tokens and Shadow Credits
  14. How Is a Skill Different from a Prompt?
  15. Action Reminders: From Candidate Advice to the Model Request
03 / 06

L3 · Agent Decisions and Execution

Entry, Main, Child, Provider, tool presentation and budgets support bounded execution.

  1. Entry: Why the Entry Point Became Simpler
  2. When Should an Agent Stop Researching and Start Answering?
  3. Provider: Why a Unified API Still Needs Shared Admission
  4. Why the Next Model Request Still Queues After HTTP 200
  5. Tool Surface: Why the Model Cannot See an Available Tool
  6. Multi-Ecosystem Hooks: Claude, Pi and Codex
  7. Checkpoint System Design: What Does an Agent Save, and What Lets It Continue?
  8. What a Model Attempt Leaves Behind: Capturing, Retaining, and Reading Agent Research Evidence
04 / 06

L4 · Tools and Domain Capabilities

Controlled reading, isolated computation, MCP, market time and investment facts; receipts are not conclusions.

  1. Web Research: From Search to Deep Reading
  2. PDF: From Text to Evidence
  3. When an Agent Says Today: User Time and Market Sessions
  4. Python: From Tool to Computational Workbench
  5. Exact Numbers Are Not Correct Investments: A Numeric Kernel for Units, Rounding, and Residuals
  6. Deterministic Fund Allocation: From Holdings to Reproducible Amount Advice
  7. Why Can’t a Portfolio Report Be Written While It Is Still Being Calculated?
  8. Connecting Portfolio Data Through a Read-Only MCP
  9. The RAG Lifecycle: From Document Ingestion to Cited Answers
  10. MCP Host: From Discovery to Usable Tools
  11. Why a Research Report Must Become an Asset Before an Agent Can Reuse It
  12. Externalized Results and Input Completeness
05 / 06

L5 · Result Publication and Delivery

Separate accepted content, committed revisions, conversation delivery and feedback facts.

  1. Publication: Why Answers Need Their Own Delivery System
  2. A Report Is Generated: Why Can’t It Be Shared Yet?
  3. When a User Dislikes an Answer: Feedback Is Not Automatic Learning
06 / 06

L6 · User Interface

Visible answers, progress and correct Run attribution; reading placement does not imply exclusive frontend ownership.

  1. An Agent Client Is More Than an SSE Renderer: Presenting Research Progress Truthfully

All articles · Architecture · Engineering

Engineering notes, evidence and boundaries. · Contact: hpoenixf@foxmail.com · WeChat: hpoenixf0510