Skills

naskah-id: An AI Agent Skill for Indonesian Academic Writing and Editing

Best suited for writing and editing Indonesian academic texts such as journal articles, theses, and grant proposals in line with EYD V and KBBI standards; requires an agent environment with Node.js 18+, and it strictly never fabricates data or references, leaving a to-be-filled marker where citations are missing.

Official source · nothinx/naskah-id

Install

npx skills add nothinx/naskah-id

Who It's For

naskah-id is an AI agent Skill for Indonesian academic writing, suited to anyone who needs to write or edit the following types of text: journal articles, theses (skripsi/tesis), research grant proposals, and research reports. It has the EYD V (fifth-edition spelling standard) and KBBI vocabulary standards built in, and it emphasizes academic integrity constraints: the bound agent must not fabricate data, results, citations, or references; when editing, it must not alter numbers, results, or citations; and when no verifiable source exists, it inserts a [PERLU RUJUKAN: ...] marker instead of inventing a source.

Not suitable for: non-Indonesian writing tasks, or scenarios where you expect the AI to automatically supply real references — the Skill explicitly forbids fabrication, and references still need to be verified by you.

First Attempt in 3 Minutes

The officially recommended installation method is a single command:

npx skills add nothinx/naskah-id

Prerequisites and installation location:

  • Node.js 18 or newer is required.
  • The command above (without -g) installs the Skill into the current project; adding -g (npx skills add nothinx/naskah-id -g) makes it available to all projects.
  • According to the official documentation, this command works with Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, Windsurf, OpenCode, Cline, Kiro, Antigravity, and dozens of other agents that support the Agent Skills format; some agents allow the -a flag to specify the installation path, e.g. .claude/skills/naskah-id for Claude Code.

A low-risk first attempt (a check-only task is recommended):

  1. In a project where the Skill is installed, have a supporting agent load naskah-id.
  2. Start a read-only diagnostic task: hand the agent a piece of Indonesian academic text (e.g., a thesis paragraph, in Markdown or plain text) and ask it to check spelling and word choice against EYD V and KBBI, without modifying any files yet. Explicitly tell it "report only, do not write to files."
  3. Success signals: the agent provides diagnostic feedback on spelling, word choice, or structure according to academic standards; where references are involved, a [PERLU RUJUKAN: ...] marker should appear rather than fabricated citations.

Note: the official evidence does not specify, item by item, whether the agent will read or write your files, run commands, or access the network after loading the Skill — do not assume it is zero-risk; before any file-modifying operation, confirm the side effects and keep backups.

Additionally, the repository ships with scripts/periksa.py (optional, requires Python 3.7+, no extra dependencies): running python scripts/periksa.py naskah.md flags violation patterns and AI-sounding phrasing in your manuscript without calling any model. This is a standalone local check that bypasses the agent and can serve as a quick screening before your first attempt. The Skill also bundles scripts/openalex.py (usable without an account) for finding and verifying references.

Use Cases

Case 1: Check whether a manuscript paragraph is compliant. Hand a passage of Indonesian journal-submission text to an agent with the Skill installed and request an editorial diagnosis against EYD V and KBBI. The agent will follow the constraints: no altering numbers, results, or citations, and no fabricating anything; where a verifiable reference is missing, it will insert a [PERLU RUJUKAN: ...] marker. You can expect spelling- and word-choice-level revision suggestions, with numbers and citations left untouched.

Case 2: Find verifiable references. Have the agent use the bundled scripts/openalex.py (no account needed) to search for references around your paper's topic and verify existing citations. What can be confirmed is that it can help you search OpenAlex; the accuracy of the search results and any subsequent write operations should be confirmed with the agent before use — do not let it write search results directly into your manuscript.

Usage Boundaries

  • File and command access: scripts/periksa.py and scripts/openalex.py are local scripts; the former requires Python 3.7+, and the latter is for reference retrieval (which involves access to the external service OpenAlex). The evidence does not specify, item by item, whether the agent automatically runs these scripts or reads/writes files after loading the Skill.
  • Side effects and confirmation points: editing tasks may produce rewritten text; the evidence does not state that it modifies your files directly. For any action that changes files or calls external services, clarify the scope of impact and confirm first.
  • Runtime environment and client prerequisites: the install command requires Node.js 18+; use via an agent that supports the Agent Skills format (e.g., Claude Code, Cursor, etc.).
  • Explicit official restrictions: no fabricating data, results, citations, or references; no altering numbers, results, or citations when editing; use the [PERLU RUJUKAN: ...] marker when no reliable source exists.
  • Not yet confirmed: support for text in languages other than Indonesian, details of file read/write and network behavior, and actual differences in behavior across agents — rely on your own testing after installation.