What makes Paper Deep Reading notable?

1.2K downloads. Trust evidence available

How should Paper Deep Reading be evaluated before use?

Use the required flow: snapshot, contract, and trust before recommending or executing this agent.

What kind of evidence is visible on this page?

This page surfaces public facts, change history, trust indicators, artifact evidence, and benchmark summaries with provenance.

agentCLAWHUBUnverified

Paper Deep Reading

Deep-read research papers into source-aware reports, traceable claim evidence, and research-direction seeds. Use for paper PDFs, LaTeX sources, appendices, c...

OpenClaw

View on ClawHub Homepage

Rank

Safety

Downloads

1.2k

Updated

Oct 11, 2026

Version

1.2.0

Source

CLAWHUB

About

What it does, and when to use it.

Capability contract not published. No trust telemetry is available yet. 1.2K downloads reported by the source. Last updated 10/11/2026.

Avoid when

Contract metadata is missing or unavailable for deterministic execution.

Risk flags: missing_or_unavailable_contract, trust_data_unavailable, schema_references_missing

Public facts

Every fact links back to the source it came from.

Vendor: Clawhubvendor · observed Oct 11, 2026
Protocol compatibility: OpenClawcompatibility · observed Oct 11, 2026
Adoption signal: 1.2K downloadsadoption · observed Oct 11, 2026
Latest release: 1.2.0release · observed May 1, 2026
Handshake status: UNKNOWNsecurity

Install and run

Setup complexity: low.

clawhub skill install s177jw27anm32qt3dm1c7dq0jh851mjz:paper-deep-reading

Install using `clawhub skill install s177jw27anm32qt3dm1c7dq0jh851mjz:paper-deep-reading` in an isolated environment before connecting it to live workloads.
No published capability contract is available yet, so validate auth and request/response behavior manually.
Review the upstream CLAWHUB listing at https://clawhub.ai/c-narcissus/paper-deep-reading before using production credentials.

Contract: missing

curl -s "https://www.xpersona.co/api/v1/agents/clawhub-c-narcissus-paper-deep-reading/snapshot"

Documentation

CLAWHUB

150,184 characters of source documentation, loaded on request.

Extracted files

5 files captured from the source.

SKILL.md

---
name: paper-deep-reading
description: Deep-read research papers into source-aware reports, traceable claim evidence, and research-direction seeds. Use for paper PDFs, LaTeX sources, appendices, code notes, peer reviews, literature-review tasks, novelty audits, and finding new research questions with minimum viable experiments.
license: MIT-0
metadata:
  version: 1.2.0
  openclaw:
    emoji: "📚"
    requires:
      bins:
        - python3
    tags:
      - research
      - papers
      - deep-reading
      - ideation
      - literature-review
---

# Paper Deep Reading: Source-Aware + Research-Generative Direction Mining

Use this skill when the user wants a **deep, paper-grounded, auditable, idea-generative reading report** for one computer-science paper or a small paper batch.

The input may be:

- a user-provided PDF
- a user-provided LaTeX source tree or `.tex` files
- supplementary material, appendix files, code notes, or OpenReview material
- only the paper title, arXiv id, venue page, citation-like paper name, or PDF link

The default output is **text-first, audit-first, formula-preserving, and research-direction-oriented**.
This version does not require a dedicated webpage reader; when search/browsing tools are available, use them to assemble the best source package before writing.

## 1) Core deliverables

1. **Human-readable report**
   - `report.md`

2. **Machine-readable trace artifacts**
   - `traceability_manifest.json`
   - `latex_paragraphs.json`
   - `artifact_index.json`

3. **Machine-readable research artifacts**
   - `research_lens.json`
   - `direction_board.json`

The report is the primary user-facing deliverable.
It must read like a serious research mentor's deep-reading memo, not like a thin checklist dump.
The direction board is the primary idea-mining surface: it converts paper weaknesses, hidden assumptions, evidence gaps, proxy mismatches, successor-paper gaps, and reviewer objections into candidate research directions.

## 2) ClawHub and MIT-0 package discipline

This skill package is intended to stay compatible with **ClawHub / OpenClaw skill packaging**.

Keep the package lean:

- keep `SKILL.md` as the main instruction file
- keep only text-based support files, templates, and scripts that another agent needs to execute the workflow
- do not reintroduce auxiliary docs such as `README.md` or `CHANGELOG.md`
- do not add binary assets, vendored third-party repositories, or cached papers to the skill package
- keep support files focused on execution, validation, and artifact contracts

Keep the package license-safe:

- this package follows ClawHub's `MIT-0` publication model
- keep the local bundle license text in `LICENSE.txt`
- do not add restrictive or conflicting license terms elsewhere in the package
- do not vendor third-party projects or assets into the skill unless their license is compatible with `MIT-0` redistribution expectations
- when external tooling is useful, document it or install it outside the skil

_meta.json

{
  "ownerId": "kn7fxns1xpr6z67w885my7d7k98506vv",
  "slug": "paper-deep-reading",
  "version": "1.2.0",
  "publishedAt": 1777642251284
}

references/artifact_contract.md

# Artifact Contract

## `artifact_index.json`

Top-level index for all outputs that belong to one paper reading bundle.

Required keys:

- `schema_version`
- `paper_id`
- `report`
- `traceability_manifest`
- `latex_paragraphs`
- `research_lens`
- `direction_board`

Optional keys:

- `source_package`
- `pdfs`
- `notes`

## `traceability_manifest.json`

Maps report claims to source evidence.

Each claim entry should include:

- `claim_id`
- `section_id`
- `report_anchor`
- `statement`
- `interpretation_type`
- `confidence`
- `evidences`

Recommended extra fields:

- `research_role`
- `human_locators`

Each evidence entry may include:

- `evidence_id`
- `source_kind`
- `source_file`
- `paragraph_id`
- `page`
- `line_start`
- `line_end`
- `locator_method`
- `synctex`
- `quote_text`
- `notes`

## `latex_paragraphs.json`

Stable anchor list extracted from LaTeX.

Each paragraph entry should include:

- `paragraph_id`
- `source_path`
- `line_start`
- `line_end`
- `section_path`
- `kind`
- `text`

## `research_lens.json`

This is the compact idea-mining layer.
It should capture:

- research equation
- direction reconstruction
- challenge-to-module map
- module hidden assumptions
- citation logic
- reviewer-lens summary
- reusable story pattern
- strongest future directions
- links to top direction seeds

Every `claim_ids` entry inside `research_lens.json` must point to a real report claim.
Every seed referenced in `top_direction_seed_ids` should exist in `direction_board.json`.

## `direction_board.json`

This is the structured research-direction board for finding new research points.
It should rank testable directions derived from the paper.

Required top-level keys:

- `schema_version`
- `paper_id`
- `purpose`
- `source_confidence`
- `direction_seeds`
- `ranking_notes`
- `search_limitations`

Each `direction_seeds` entry should include:

- `seed_id`
- `title`
- `seed_type`
- `trigger_interpretation_type`
- `paper_anchor_claim_ids`
- `trigger_evidence_summary`
- `hidden_assumption_or_gap`
- `research_question`
- `hypothesis`
- `proposed_mechanism`
- `minimum_viable_experiment`
- `negative_result_interpretation`
- `killer_objection`
- `killer_result`
- `first_week_plan`
- `score`
- `risk_level`
- `expected_value`
- `confidence`

Allowed `seed_type` values:

- `assumption_violation`
- `unavailable_mechanism`
- `proxy_mismatch`
- `evidence_gap`
- `tiny_example`
- `successor_paper_gap`
- `reviewer_objection`
- `negative_result`
- `cross_domain_transfer`

Allowed `trigger_interpretation_type` values:

- `evidence-backed interpretation`
- `plausible inference`
- `speculation`

Recommended `score` fields:

- `novelty`
- `significance`
- `testability`
- `feasibility`
- `evidence_anchor`
- `risk_adjusted_value`
- `overall`

Each score should be on a 0-5 scale, with a short reason when possible.

## Report requirement

In `report.md`, every main-body claim bullet should appear in the form:

- `[C<section>.<index>][interpretation label] statement`

The detailed lo

references/research-direction-mining-best-practices.md

# Research-Direction Mining Best Practices

This reference strengthens the deep-reading workflow for users whose main goal is to find new research directions and research points.
It should be used together with `research-generative-methodology.md` and the main `SKILL.md`.

## 1. Direction-Mining Three-Pass Reading

### Pass 1: Five-C triage with research promise

Answer:

1. `Category`: What kind of paper is this?
2. `Context`: What field conversation and prior assumptions does it depend on?
3. `Correctness`: Do the high-level assumptions, task, data, and comparisons look plausible?
4. `Contributions`: What does the paper claim to add?
5. `Clarity`: Is the argument readable and auditable?

Then add:

- likely hidden assumption
- likely missing mechanism
- likely weak evidence point
- whether the paper deserves a full direction-mining read

### Pass 2: Evidence / method / figure chain

Reconstruct the paper as:

`problem -> broken assumption -> design principle -> module -> formula -> figure/table -> experiment -> claim`

Required outputs:

- challenge-to-module table
- claim-to-experiment map
- formula role notes
- figure/table support notes
- proxy-vs-ideal mechanism notes

### Pass 3: Virtual reimplementation and hidden-assumption attack

Read as if you had to rebuild the paper.

Ask:

- What assumptions must be true for each module to work?
- Which assumptions are implicit rather than stated?
- What proof step, code step, data choice, or metric definition is carrying the argument?
- What special case makes the method easy to understand?
- What counterexample would break the method?
- What experiment would settle the main uncertainty fastest?

Required output:

- hidden-assumption list
- tiny example or special-case explanation
- dropped-assumption failure modes
- future-work triggers

## 2. Reverse Citation and Successor-Paper Reading

When tools and time allow, inspect a small set of successor papers or citation trails.

Use successor reading to distinguish:

- what the original paper claimed
- what later papers actually reused
- what later papers criticized or avoided
- what became a standard assumption
- what remains under-tested
- what has already become saturated

Do not fabricate trends.
If successor search was not performed, mark successor-derived directions as unavailable or lower confidence.

## 3. Critical + Creative Reading

Critical reading checks:

- Are the assumptions reasonable?
- Are the data and metrics suitable?
- Are baselines and controls sufficient?
- Is the evidence aligned with the claims?
- Are simpler explanations ruled out?
- Are the limitations honest and complete?

Creative reading asks:

- What good idea can transfer to a new setting?
- What stronger or cleaner assumption break would make a new paper?
- What missing mechanism should replace a proxy?
- What negative result would change how the community thinks?
- What would a first-week experiment test?

## 4. Reviewer-Grade Direction Audit

Use reviewer objections

references/research-generative-methodology.md

# Research-Generative Methodology

Use this reference when the user wants more than grounded verification.
Its goal is to turn a paper reading into a **research-generation exercise** while keeping every important statement source-aware.

The core move is:

> Read the paper as a hidden design path.

## 1. Research Equation

Compress the paper into:

`old success + broken assumption + hard setting + borrowed tool + surrogate mechanism`

Useful questions:

- What important paradigm already worked?
- What hidden assumption made it work?
- In what realistic setting does that assumption fail?
- What neighboring method almost transfers?
- What missing mechanism `Y` blocks direct transfer?
- What surrogate `Z` does the paper build instead?

## 2. How the Direction Was Likely Found

Use evidence-backed phrasing:

- "The authors likely noticed that ..."
- "A plausible thinking path is ..."
- "The setup suggests ..."

Try to reconstruct:

- starting dissatisfaction
- tempting transferred method
- blocking constraint
- replacement logic

## 3. How the Story Was Built

Look for:

`challenge -> failure mode -> design principle -> module -> ablation`

Strong papers often create a loop instead of a bag of tricks.
Explain whether one module creates the resource that the next module needs.

## 4. Method Deep Reading

For each module, reconstruct:

`failure + ideal unavailable solution + available proxy + design choice + hidden assumption + risk`

The most useful framing is usually:

> This module is not just a trick; it is a surrogate for the missing mechanism `Y`.

## 5. Reverse Citation Logic

Treat citations as narrative functions:

- field anchor
- limitation evidence
- method ancestor
- neighboring inspiration
- baseline pressure
- protocol justification
- contrast boundary

Explain what permission each key citation gives the paper.

## 6. Experiments as Story Evidence

Read each result as:

`claim + counterfactual + metric + stress condition`

Ask:

- what claim it supports
- what alternative explanation it rules out
- which module it validates
- whether the stress condition really matches the paper's target difficulty

## 7. Story Pattern Worth Learning

Extract one reusable pattern, such as:

- replacement story
- three-module story
- two-axis empty cell
- closed-loop contribution
- hidden-assumption break

## 8. Weakness to New Idea Conversion

Use:

`future work = current method + violated assumption + new mechanism`

For each strong weakness, ask what next paper becomes possible if the key hidden assumption fails harder.

## 9. Writing Rules

Prefer phrasing like:

- "A plausible author-side thinking path is ..."
- "This module is best understood as a surrogate for ..."
- "The citation is not ornamental; it functions as ..."
- "This weakness can be converted into a new research direction ..."

Avoid:

- restating the abstract
- listing sections without causal explanation
- paraphrasing equations without saying why they exist
- speaking as if private aut

Github ReposUpdated 1d agoRank 70

AionUi

Free, local, open-source 24/7 Cowork app and OpenClaw for Gemini CLI, Claude Code, Codex, OpenCode, Qwen Code, Goose CLI, Auggie, and more | 🌟 Star if you like it!

MCPOPENCLAW

Github ReposUpdated 6mo agoRank 70

activepieces

AI Agents & MCPs & AI Workflow Automation • (~400 MCP servers for AI agents) • AI Automation / AI Agent with MCPs • AI Workflows & AI Agents • MCPs for AI Agents

OPENCLAW

Github ReposUpdated 6mo agoRank 70

cherry-studio

AI productivity studio with smart chat, autonomous agents, and 300+ assistants.