AI agent skill · by @ARA-Labs

Research Foresight

ARA World Model — read-only reasoning engine over ONE Agent-Native Research Artifact (ARA), run LOCALLY with the coding agent itself as the LLM (no SDK, no API key). Given an ARA directory and a free-text query, it answers any question about the ARA — a forward "what if I change X", but equally why-did-this-work, what-should-I-try, is-this-sound, how-do-these-compare, or anything else — by retrieving precedent from the ARA's native files (references/RETRIEVE.md) and answering as the Predictor (references/PREDICT.md): a bold, grounded, falsifiable Answer shaped to what the question actually calls for. TRIGGERS: ask the world model, wm predict, predict with the world model, what if I change X, forecast the loss curve, will this help, why did this work, what should I try next, is this claim sound, compare these, retrieve precedent, what precedent surfaces.

Archived 2026-10-23 · Source updated 2026-10-11 · License: MIT. Check the original repository for the latest version.

SKILL.md instructions

# research-foresight — the ARA World Model

You (the coding agent) are the LLM that runs the engine — no SDK, no API key, no network call. The
engine is three reference contracts under this skill's `references/` directory (quote every path;
it may contain spaces):

- `references/CONTRACT.md` — the foundation both contracts bind to; if documents disagree, it wins.
- `references/RETRIEVE.md` — the Retriever: agentic search + semantic rank over the ARA's native files.
- `references/PREDICT.md` — the Predictor: grounded, honest answering of the question asked.

## Inputs

From the user's message (or `$ARGUMENTS`): an **`<ara_dir>`** (the ARA in scope) and a free-text
**query**. If `<ara_dir>` turns out not to be an ARA (a plain paper, repo, or notes folder),
compile it into one first with `/compiler <path>`, then rerun this skill.

## Procedure

1. **Retrieve — adopt `references/RETRIEVE.md`.** Read it now and follow it exactly against
   `<ara_dir>`.
2. **Answer — adopt `references/PREDICT.md`.** Read it now and follow it exactly, consuming the
   retrieval from Step 1.

Render the `answer` prominently, then the honesty envelope (`grounded_inference` /
`speculative_leap` / `basis` / `reasoning` / `confidence` / `confidence_reason` / `falsifiable`).

The engine is read-only: read nothing outside `<ara_dir>` and this skill's `references/`; write
nothing anywhere.