skills.rest
Immortalqx
ImmortalqxCOMMUNITY
Explore all skills by this author →
WebsiteGitHub

research-survey-loop

Runs multi-round literature surveys with task state files, chunked PDF reading, and incremental Chinese survey writing.

52|Updated Apr 24, 2026
Search Tags:#pdf extraction#semantic scholar#robotics#arxiv#embodied ai#research workflow#literature survey
One-click install
npx skills add https://github.com/Immortalqx/my_codex_skills --skill research-survey-loop-immortalqx
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: research-survey-loop
Source: https://github.com/Immortalqx/my_codex_skills/tree/main/research-survey-loop
Command: npx skills add https://github.com/Immortalqx/my_codex_skills --skill research-survey-loop-immortalqx

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) and assets (resource) components.

What problem does it solve? Long-running literature reviews lose context between sessions: search priorities drift, local PDFs pile up in an unsorted pool, and notes become disconnected per-paper cards. This Skill maintains a stable task directory with a fixed contract (task.md, round_log.md, current_task.md, survey.md) so every round resumes from a precise entry point and the survey grows incrementally. ## Core Features & Use Cases - Persistent task state: Bootstraps a survey_tasks/<topic-slug>/ directory from templates and rewrites current_task.md each round with explicit search, reading, and migration queues. - Prioritized source retrieval: Searches Nature/Science, top CV venues (CVPR/ICCV/ECCV/TPAMI/IJCV), top robotics venues (ICRA/CoRL/RSS/IROS), then arXiv via Semantic Scholar and arXiv APIs, downloading PDFs into task-local sources/papers/. - Disciplined reading and writing: Extracts PDF text in windows of at most 10 pages, migrates absorbed local PDFs out of the root papers/ pool, and grows one long Chinese survey.md by category paragraphs with relative-path citations. - Use case: Ask for a survey on world models for robot navigation; the Skill creates the task directory, searches publisher-first sources, reads papers in chunks, and appends categorized Chinese paragraphs round after round. ## Quick Start Use the research-survey-loop skill to start a long-running survey on embodied AI navigation and run the first search and reading round.

Frequently Asked Questions about research-survey-loop

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

FAQPage Schema
How do I run a multi-round literature survey that persists between sessions?▼

Initialize a task directory with init_task.py, which creates task.md, round_log.md, current_task.md, and survey.md from templates. Each round appends to round_log.md and rewrites current_task.md so the next session resumes from explicit search and reading queues.

How do I search arXiv and Semantic Scholar for survey papers from the command line?▼

Run fetch_sources.py search with a query and task directory to get normalized JSON results from both the Semantic Scholar Graph API and the arXiv API. Add --download-top N to automatically download the first N results that expose an arXiv ID or open-access PDF.

How do I read a long PDF in chunks for a literature review?▼

Use extract_pdf_chunk.py with --start-page and --end-page; it enforces a maximum of 10 pages per call using pdfinfo and pdftotext. The JSON output includes the next page window so long papers are read in consecutive windows like 1-10, 11-20, 21-30.

What research topics does this survey workflow cover?▼

The workflow targets robotics, embodied AI, computer vision, world models, spatial intelligence, navigation, manipulation, and 3D scene understanding. Its venue priorities and templates are optimized for robotics and CV communities, though adjacent computer science topics can be surveyed.

What happens when Semantic Scholar or arXiv is unavailable during a search?▼

The search script records a warning for the failed source and continues returning results from the remaining APIs. The workflow rules also direct the agent to keep working with other sources rather than blocking the round.

Can I reuse papers already downloaded in another survey task?▼

Yes, fetch_sources.py reuse-task-paper copies PDFs from another task's sources/papers directory into the current task. Root papers/ files are moved into the task only when actually absorbed, and every migration is recorded in round_log.md.

Related Skills

Explore More Education & Research Skills→
HeXiao-55HeXiao-55
1

arxiv

Search arXiv by topic or ID, download PDFs, and summarize metadata.

Community
Intermediate
u9401066u9401066
11

concept-development

Validate research concept drafts and assess novelty with reviewer-style critique.

Community
Advanced
kridaydavekridaydave
1

beacon

Surveys research literature to map prior work, state of the art, and open gaps.

Community
Intermediate
Shallow-WShallow-W

alphaxiv

Retrieve arXiv paper summaries with progressive fallback to LaTeX source.

Community
Intermediate