researchforge-papers · git:20260713.896358d · 2026-07-13 · sha256 5bf748ffac008afd

researchforge-papers git:20260713.896358dA

Immutable. This exact content never changes and is served at /api/v1/blob/5bf748ffac008afd.

No tool to install first
claude plugin marketplace add https://markdownregistry.com/mp/a/art_meo6fvrpotdrlyk3/5bf748ffac008afd8e470a92d7b26c62b17bb0573b16fcfbed46c2ad5f0126dd.json

Then /plugin install researchforge-papers@markdownregistry-meo6fvrpotdrlyk3 in Claude Code. It installs this exact version, the one graded A here, pinned to the commit it was graded at, and leaves it there.

---
name: researchforge-papers
description: Search arXiv for papers relevant to the project objective and review what was stored. Use when the user wants literature, related work, or asks what papers ResearchForge found.
---

# Find and review papers

## Search

The engine generates queries from the objective (and repository scan, if
any), retrieves candidates from arXiv, deduplicates, ranks, and stores them:

```bash
researchforge research search --json
```

To steer the search, pass explicit arXiv queries (repeatable):

```bash
researchforge research search -q "all:mixture of experts routing" --json
```

Report how many papers were retrieved, deduplicated, and stored — from the
JSON, not from memory.

## Review

```bash
researchforge papers list --json
researchforge papers show <paper-id> --json
```

When discussing a paper, use only its stored metadata (title, authors,
abstract, ranking signals). ResearchForge stores metadata and abstracts, not
full texts — say so if the user asks for details an abstract cannot support.

Next step after papers exist: the researchforge-landscape skill synthesizes
them into research directions.

## Rules

- The Python engine is the boundary: never work around a validation error, a
  protected path, or an approval gate — fix the artifact or ask the user.
- Approvals belong to the user: never pass `--yes` or type a confirmation
  unless the user explicitly approved that step in this conversation.
- Ground every summary in stored data: quote only numbers returned by
  `--json` output or files under `.researchforge/` — never invent metrics.