paper-reading-skills

Paper Reading Skills

Languages: English | Chinese

A human-in-the-loop Codex skill suite that turns paper reading into staged, auditable research decisions.

Paper Reading Skills is a suite of Codex skills for early-stage research. It helps researchers organize paper reading around a clear research motivation at each stage and turn reading results into durable artifacts that can support the next research move.

Paper Reading Skills is built around staged confirmation. The agent proposes boundaries, paper sets, candidate questions, method needs, close works, claims, risks, and next steps; the researcher confirms, revises, rejects, or explicitly delegates consequential decisions before the workflow moves on. The intended outcome is that the researcher understands the project deeply enough to own the next decision.

In this loop, the agent broadens the search space, structures options, and challenges weak assumptions; the researcher learns the project, calibrates the direction, and supplies the judgments that only a human owner can make.

Efficient paper reading starts by clarifying these questions before reading:

• Why am I reading these papers?
• Which research decision am I trying to make right now?
• Which parts of the papers should I focus on?
• What should this reading session produce?
• When is it enough to stop?

This project breaks paper reading into a set of common research actions, with a workflow, input boundary, reading strategy, output format, and stopping condition for each action.

---

Core Idea

Ordinary paper reading often aims at "finishing papers." This project aims at "making research judgments."

For example:

• When entering a new direction, start by building a field map instead of rushing into paper summaries.
• When you have an early idea, run a problem reality check before designing a method.
• When a method starts taking shape, pass through method commitment before treating it as a final plan.
• When preparing to write a paper, organize around close work, claim support, and positioning instead of broad related-work collection.

Each skill corresponds to a concrete research stage and produces artifacts that downstream workflows can consume.

---

What Makes This Different

Paper Reading Skills treats research as a collaborative decision process, not a fully automated content-generation task.

• Stage gates keep important choices visible: field boundaries, seed papers, method needs, candidate methods, experiment claims, close works, and final commitments are reviewed before they drive downstream work.
• The agent is an active research partner: it searches, structures, compares, challenges, and recommends, but it does not silently turn its own recommendation into the researcher's commitment.
• The researcher stays inside the loop: they learn the project while making decisions, correct the agent's framing, and provide judgment where literature evidence alone is not enough.
• Delegation is recorded rather than disguised as confirmation. If the researcher asks for a more automatic run, the output stays provisional where key gates were not reviewed.

This is meant to create research ownership. By the end of a workflow, the researcher should be able to explain the boundary, evidence, risks, alternatives, and next move, rather than merely receive an impressive document about a topic they still do not know.

---

Skill Map

Stage	Skill	Research Question It Helps Answer	Main Output
01	Field Map	What are the main branches, method routes, and research opportunities in this field?	Field map / research clusters / opportunity candidates
02	Research Question	Which gaps can become actionable research questions?	Research question cards
03	Problem Reality Check	Is this problem real, or only superficially plausible?	Problem justification / risk notes
04	Theoretical Grounding	Is there theoretical support that can strengthen the problem framing?	Theoretical grounding package
05	Method Inspiration	Which mechanisms from papers can be transferred into candidate methods?	Method candidate library
06	Method Commitment	Is the current method stable enough for the researcher to own?	Committed / provisional / redesign / rejected method outcome
07	Experiment Design	How can claims be translated into verifiable experiment designs?	Experiment plan / baseline pressure matrix / claim-metric map
08	Research Framing	How should this research be positioned in the paper narrative and related-work comparison?	Research framing package
Extra	Workspace Presentation	How can an existing research workspace become a browsable, handoff-ready interface?	Interactive presentation workspace

---

Output Modes