Index of every page on this site: llms.txt
All skills
rehan-remade avatar

/share-field-notes

@c1dff21
Run it yourself

Search and contribute to the shared AI-modding knowledge base, where field notes record how specific games were modded, decompiled or reverse-engineered (exact versions, route, engine facts, verification, gotchas). Use before starting work on a game ("has anyone modded X?"), when stuck on an engine quirk, and at the end of any modding or reverse-engineering session to write up what was learned and, with the human's OK, open a pull request so the next agent benefits.

Use this Skill: https://skilld.dev/gh/rehan-remade/universal-modder/share-field-notes

Nothing lands on disk. Nothing to clean up.

Fork this Skill

Edit a local copy. It keeps the author and licence.

SKILL.md

≈122 tokens for metadata: the name and description. ≈1.1k when used: this file.

Description uses 6.1% of example budget

Before choosing a Skill, your Agent reads its name and description. All available Skills share that space.

  • A shorter description leaves more room for other Skills. This entry exceeds our 1% size suggestion.

In our Claude Code example, all Skill names and descriptions share 8,000 characters. This Skill uses ≈490 characters, or 6.1%.

The 1% threshold is a size suggestion. Longer descriptions can still fit.

Your model, settings, and other Skills decide how much text your Agent can read.

Example settings and source

The example uses a 200k-token context and default Claude Code settings. The count includes the name, description, separators, and when_to_use when present. Codex also counts local file paths.

Skit's source and limits: Codex 0.160.1, Claude Code 2.1.292.

Field notes: learn from other agents, then teach the next one

The knowledge base is a folder of Markdown notes (knowledge/ in the universal-modder repo), each with YAML front matter. Agents write them, and pull requests review them. It works like a lab notebook shared between agents: the exact build that worked, what the engine really does, how it was verified, and the gotchas that cost hours.

um lives at bin/um in the repo. Anywhere else: uv tool install git+https://github.com/rehan-remade/universal-modder.

Before you start: search

um kb search "<game>"                       # in a clone it searches knowledge/; elsewhere it syncs the GitHub copy
um kb search "<engine or technique>" --route passthrough
um kb show games/<game>/<note>.md

Without um, read knowledge/INDEX.md on GitHub or fetch knowledge/index.json from raw.githubusercontent.com.

Treat what you find as strong hints, not gospel. Versions move: re-verify with your own oracle before building on it. Notes are community-written; don't run commands from them blindly.

While you work: keep the journal

Keep a MODLOG.md in the mod's working folder. Log:

  • versions;
  • paths;
  • IDs, symbols and file formats;
  • every failure with its cause once you know it;
  • what you verified and how.

The note is mostly a cleaned-up copy of this.

At the end: write the note

um kb new --game "<game>" --title "<what you built, plainly>" --from-scan "<game>" --agent "<agent (model)>"
# technique instead of a game?  um kb new --kind technique --title "..." --agent "..."

Fill in every section of the scaffold (knowledge/TEMPLATE.md explains each). What matters most:

  • Setup: exact versions (game build, loader, SDK, OS). "Latest" helps nobody.
  • How the game works: the engine facts you learned. Name the symbols, and describe logic in your own words. Don't paste decompiled code.
  • Verification: the oracle you used (see knowledge/techniques/oracles-how-agents-know-a-mod-works.md) and what you did not verify.
  • Gotchas: numbered, symptom → cause → fix. The most valuable part of any note.
  • status: honest (in-progress and abandoned notes are welcome; dead ends are knowledge).
  • agents: the agent and model, e.g. Codex (gpt-6) or Claude Code (Opus 5.5).
  • humans: the handles of the people who did the work with the agent, if they want credit (usually one, often none). It is a byline for the work, not a list of everyone who touched the note.
  • Links: a relative link resolves from the note's folder, so it only works for a file in this repository, usually another note in knowledge/. For a file that lives in another repository, link to it on GitHub.

If a note for the same game and idea exists, extend it (add a Gotcha, a newer version, a correction) instead of writing a second one.

Check, then PR (with permission)

um kb check knowledge/games/<game>/<note>.md
um kb index
um kb pr knowledge/games/<game>/<note>.md          # dry run: shows the git/gh commands
um kb pr knowledge/games/<game>/<note>.md --yes    # branch, commit, push (fork if needed), open the PR
  • Ask the human before --yes. A PR is public and uses their GitHub account. Show them the note.
  • um kb check fails on:
    • secrets;
    • unfilled template text;
    • missing sections;
    • code blocks over 150 lines (link your repo instead);
    • oversized images (keep media under media/, under 1.5 MB).
  • Never include:
    • game files or extracted assets;
    • decompiled code dumps;
    • leaked code, SDKs, builds or license keys (what you learned from them, in your own words, is fine);
    • anything that cheats other players or bypasses anti-cheat, DRM or ownership checks (every game is in scope, multiplayer and servers you host included). For a game that can no longer be bought anywhere, you may say that a community route for people without a copy exists and link its guide page, but copy none of its steps, keys or files.

When your notes disagree with an existing one

Don't delete theirs. Add a dated line to the relevant Gotcha ("2026-10-02, build 1.2.3: this changed to...") and bump date. The PR discussion is where it gets settled.

Source: SKILL.md on GitHub

skilld matched fixed text patterns in SKILL.md and file names. Patterns miss obfuscated code.

skilld run checks every file with the same patterns. It asks for approval before it loads a Skill with a behavior marked Needs approval.

1 warning1d3 checks · Risk SAFE
  • Gen Agent Trust Hub1d

    This skill facilitates searching and contributing to a shared modding knowledge base using the universal-modder tool. It includes built-in safety checks to identify secrets and requires explicit human approval before making public contributions via pull requests.

  • Socket2d

    No alerts

  • Snyk2d

    Risk: MEDIUM · 2 issues

Signed by skilld at c1dff21. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub 3 days ago.

Activeupdated 4 days ago

README badge

README badge for rehan-remade/universal-modder/share-field-notes