Skillget

academic-writing-toolkit

agent
Security Audit
Fail
Health Warn
  • License — License: MIT
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Low visibility — Only 5 GitHub stars
Code Fail
  • rm -rf — Recursive force deletion command in scripts/doctor.sh
  • rm -rf — Recursive force deletion command in scripts/repair.sh
  • rm -rf — Recursive force deletion command in scripts/test.sh
Permissions Pass
  • Permissions — No dangerous permissions requested
Purpose
This toolkit provides structured, local agent skills to help researchers read academic PDFs, take notes, verify references, and export documents. It works as a set of offline prompts and helper scripts compatible with AI coding runtimes like Claude Code, Codex, and Cursor.

Security Assessment
Overall risk is rated as Medium. The tool does not request dangerous system permissions, and no hardcoded secrets or obvious network exfiltration routes were found. However, the automated scanner flagged multiple `rm -rf` (recursive force deletion) commands inside the repository's shell scripts (`scripts/doctor.sh`, `scripts/repair.sh`, and `scripts/test.sh`). While these deletion commands are likely intended to clean up temporary folders during routine operations, the presence of aggressive disk deletion in automated scripts is a notable security risk. Developers should manually review the exact directory paths targeted by these scripts before running `make setup` or executing any shell files.

Quality Assessment
The project uses the highly permissive MIT license and is actively maintained, with repository updates pushed as recently as today. The project layout is logical, well-documented, and clearly explains its intended workflow for academic writing. However, community trust and visibility are currently very low. With only 5 GitHub stars, it is likely an early-stage or niche project that has not yet been extensively peer-reviewed by the broader open-source community.

Verdict
Use with caution — the concept is useful and actively maintained, but you should review the shell scripts for dangerous deletion commands before running them on your local machine.
SUMMARY

Structured skills for reading, writing, and managing academic research with AI agents. Works with Claude Code, Codex CLI, Gemini CLI, Cursor, and more.

README.md

academic-writing-toolkit

Structured local agent skills for academic reading, writing, reference checking, and export.

License: MIT
Agent Skills

This repository is a public toolkit. It contains reusable skills, scripts, templates, and tests only. Put your own chapters, PDFs, notes, and private project material in your clone.

/read -> /note -> /map -> /integrate -> /audit -> /style -> /logic-review -> /export
             |                         |
             v                         v
        /verify                  /verify-refs
             |
             v
        /progress

Installation

Use git clone, not GitHub's "Download ZIP". This repository uses symlinks under .agents/skills/ so Codex, Gemini, and other local agent hosts can discover the same skills as Claude Code.

git clone https://github.com/yha9806/academic-writing-toolkit.git my-writing-project
cd my-writing-project
make setup

Then open the folder in a local agent runtime:

Runtime Local discovery path
Claude Code .claude/skills/
Codex .agents/skills/
Gemini CLI .agents/skills/
Cursor .cursor/rules/ baseline rules

The intended use is the same as local Superpowers-style skills: install the toolkit into a project folder, let the agent discover the skills, and drive the workflow from natural language or slash commands.

Skills

Skill Purpose
/read Read academic PDFs page by page with structured output
/note Record reading notes in the shared notes format
/verify Fact-check factual claims during reading
/map Map sources to chapters and coverage gaps
/integrate Propose and apply approved note-to-chapter integrations
/audit Check numbers, terminology, cross-references, and citations
/style Check and safely fix British English spellings
/logic-review Review paragraph flow, transitions, and argument continuity
/verify-refs Validate BibTeX reference records offline, with room for explicit online checks
/progress Show reading, writing, and coverage progress
/export Convert Markdown outputs to .docx and ZIP packages

Detailed guides live in docs/skills/.

Project Layout

my-writing-project/
├── .claude/skills/          Claude Code skills
├── .agents/skills/          Symlinks for Codex, Gemini, and compatible agents
├── .cursor/rules/           Cursor baseline rules
├── chapters/                Your chapter drafts
├── literature/
│   └── reading_notes/       One notes file per source
├── final_output/            Exported documents
├── scripts/                 Deterministic helper checks
├── CLAUDE.md                Canonical project configuration
├── AGENTS.md                Generated local-agent configuration
└── GEMINI.md                Generated Gemini configuration

Configuration

Edit CLAUDE.md for project-specific settings, then run:

make sync

AGENTS.md and GEMINI.md are generated from the shared block in CLAUDE.md; do not edit them directly.

Key settings include:

  • chapter, literature, notes, and export directories
  • reading page limits
  • British English writing policy
  • citation style (harvard, apa, chicago-author-date, mla, ieee, vancouver, gb-t-7714-2015)

Testing And Maintenance

make doctor   # read-only environment check
make repair   # repair symlinks and generated configs where possible
make test     # full regression suite
make sync     # regenerate AGENTS.md and GEMINI.md

The regression suite covers local skill discovery, config sync, export assumptions, citation auditing, public-content cleanup, British English checks, paragraph-logic checks, and offline reference verification.

Reference Verification

/verify-refs uses a deterministic offline core by default:

  • parse BibTeX from .bib files or Markdown code fences
  • check required fields by entry type
  • detect duplicate keys
  • validate DOI, URL, and arXiv identifier shape

When an agent has explicit network permission, the same workflow can use CrossRef, Semantic Scholar, and arXiv as external metadata sources. Project-specific self-citation rules are intentionally excluded from this public toolkit.

License

MIT. See LICENSE.

Reviews (0)

No results found