Agent skill

ResearchForge ships an agent skill pack. Agents can install it from GitHub, use rforge, run literature searches, record provenance, expand citations, screen/extract evidence, and build auditable review artifacts.

Ask your agent

Read https://github.com/TrebuchetDynamics/research-forge/blob/main/SKILLS.md,
install the ResearchForge skills, then use rforge to research: <your topic>

This works as a generic instruction for Claude, Pi, Codex, OpenCode, and similar coding agents. If the harness supports local skills, use the install command below.

Install the skill pack for Claude Code or Pi

tmp=$(mktemp -d) && \
  git clone --depth 1 https://github.com/TrebuchetDynamics/research-forge "$tmp" && \
  mkdir -p ~/.claude/skills && \
  cp -R "$tmp"/skills/* ~/.claude/skills/ && \
  cp "$tmp"/SKILLS.md ~/.claude/skills/research-forge/SKILLS.md && \
  rm -rf "$tmp"

The copied SKILLS.md is the map. Each skills/*/SKILL.md file is an installable agent skill.

What the pack includes

Research workflow

Main research-forge skill for topic research, provenance capture, citation expansion, and review packaging.

Source ingestion

Scholarly source search, imports, dedupe, source refs, and connector-aware workflows.

Documents and screening

Legal PDF acquisition, parsing, passage extraction, include/exclude screening, and PRISMA-ready records.

Evidence and analysis

Extraction schemas, source-backed evidence tables, meta-analysis runs, reporting, and package replay checks.

Use it

Use the research-forge skill to research: <your topic>

Good agent runs still require source terms review, owner approvals for scientific claims, and explicit redaction before sharing packages.