Skillget

wallaby-agent-rules

agent
Security Audit
Pass
Health Pass
  • License — License: MIT
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Community trust — 12 GitHub stars
Code Pass
  • Code scan — Scanned 1 files during light audit, no dangerous patterns found
Permissions Pass
  • Permissions — No dangerous permissions requested

No AI report is available for this listing yet.

SUMMARY

AI remembers, you find. Long-term memory + project map for any AI coding tool — two paste-in prompts, five plain files.

README.md

wallaby-agent-rules

License: MIT
PRs Welcome

AI remembers, you find.

Give your project's AI a long-term memory — and give yourself a map of the project. Two paste-in prompts, five plain files, zero dependencies. Works with any AI coding tool.

The 30-second version

Before After
Fifty messages in, the AI's memory starts to rot — settled decisions get relitigated, file names get invented, and you stop trusting it. MEMORY.md: one dated, sourced line per fact. On any conflict the file wins — flagged to you, never silently resolved.
"Where was that file?" — two weeks later, nobody knows. INDEX.md: one line per file, always current. A lookup, not a search.
The project root slowly fills with scratch and build leftovers. python3 scripts/health_check.py flags the clutter in seconds, weekly.

Get started — pick a path

Both paths are one prompt you paste to the AI working in your project. Full text lives in PROMPT.md.

🚀 One-click, ~30 seconds — all defaults (solo developer), zero questions:

Fetch https://raw.githubusercontent.com/Dawncoral/wallaby-agent-rules/main/PROMPT.md
and follow its "L0 — One-click install" section in this project.

🎯 Custom, ~2 minutes — the AI asks you 7 short questions (tool, project size, solo/team, chat volume, record style, tidy-up on/off, your ritual words), then builds the same system tuned to your answers:

Fetch https://raw.githubusercontent.com/Dawncoral/wallaby-agent-rules/main/PROMPT.md
and follow its "L1 — 6-question interview" section in this project.

Can't fetch URLs? Open PROMPT.md and paste the section directly — it's plain text.

What you get

File Job
AGENTS.md (or CLAUDE.md / .cursorrules, per your tool) Entry point — tells the AI to read its memory every session
MEMORY.md Long-term memory: permanent facts, decisions, iron rules
NOW.md Current state: work in flight, recently touched
INDEX.md The project map: one line per file — first version generated by scanning your project on the spot
scripts/health_check.py Zero-dependency weekly check, seven scans: stray root files, build artifacts, index drift both ways, stale entries, entry-file size walls (bytes/lines/tokens, CJK-aware), undated facts

The moment installation finishes, you watch the AI scan your project and hand you the first INDEX.md — "you find" delivered on the spot.

New in v3: the closeout ritual. Say "wrap up" and the AI triages everything the session produced — into long-term memory, the dated log, current state, or the bin — then checks what drifted from the plan. Memory stays true because someone closes. See the L3 prompt in PROMPT.md.

How it works (three sentences)

  1. The entry file makes the AI read MEMORY.md + NOW.md before every session — memory is a habit, not a feature.
  2. INDEX.md turns "find me X" from a search into a lookup, and every new file is registered the moment it's created.
  3. A weekly health check catches drift — stray files, artifacts, unregistered docs — before the mess compounds.

Everything is plain Markdown you can read, git diff, and edit. No vector store, no service, no lock-in.

Updating

Already using this? Paste the L2 — Upgrade check prompt from PROMPT.md to your AI. It detects your version (via the <!-- wallaby-agent-rules vX --> marker or file fingerprints), then proposes an incremental upgrade list — you approve item by item. Your content is never overwritten. Version history and migration steps: CHANGELOG.md.

Roadmap

One line: v3 shipped the closeout ritual and the deeper health check; next open drop is multi-agent / team memory discipline. Deeper governance (architecture review, managed oversight) is planned as a hosted offering rather than an open drop — watch the repo.

Also in this repo

  • templates/ — NOW / INDEX / LOG templates + the health check script, ready to copy.
  • AGENTS.md / MEMORY.md — the v1 starter templates (token-budget discipline, three-tier memory, stakeholder register), still valid and still dogfooded by this repo itself.
  • examples/ — filled-in MEMORY / NOW / INDEX examples: what "good" looks like two weeks in, not just on day one.

Who built this

Wallaby Token (WALLABY DATA PTY LTD, ABN 90 701 729 964) provides inference-as-a-service for open-weight large language models. This memory system runs our own project daily — not advice we sell, but infrastructure we depend on. Full write-up: wallabytoken.com.

License

MIT — adapt freely, attribution appreciated.

Reviews (0)

No results found