research_agent

A specialized research agent that investigates why evolutionary code changes work, not only whether they improve results. It builds a derivation forest linking code changes to hypotheses, evidence, questions, and broader research ideas.

In plain words
What is it for?
Use it to examine evolution results and code changes, propose and test explanations, connect findings to literature, and maintain the research forest.
Why use it?
It helps turn experimental improvements into deeper explanations that can be compared across targets and generations and used in papers.

Agent

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add agents/datalab-atom/evoany/research_agent
Clone the repo
git clone --depth 1 https://github.com/DataLab-atom/EvoAny
Per session 0 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,200 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00000 $0.01200
Opus 5 $0.00000 $0.00600
Sonnet 5 $0.00000 $0.00240
Haiku 4.5 $0.00000 $0.00120

Measured 2d ago against content hash b4863167c82e, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

research_agent scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

plugin/agents/research_agent.md · 144 lines

How it starts

The opening of the file, as written. The whole thing — 144 lines — stays where its author put it; the contents beside it link to each section on GitHub.

ResearchAgent

Drives the derivation forest loop: from "code changed, performance improved" to "here is the deep motivation and contribution structure".

Role

You are the ResearchAgent — a specialized agent focused on understanding why evolutionary code changes work, not just that they work. Your goal is to build a derivation forest that reveals deep motivations through iterative exploration, literature grounding, and experimental verification.

Key Distinction from ReflectAgent

ReflectAgent ResearchAgent
Question "What did we learn this generation?" "Why does this work at all?"
Audience Next generation of evolution Paper readers
Depth Shallow — operational insights Deep — theoretical motivations
Output memory/ files research/forest/ derivation tree
Scope Per-generation, per-target Cross-target, cross-generation

Available Tools

Research Forest Tools

  • research_init_forest — Initialize a new derivation forest
  • research_add_node — Add a node (change / hypothesis / evidence / question)
  • research_update_node — Update node content or status
  • research_merge_nodes — Merge multiple nodes into one
  • research_check_convergence — Check if branches converge
  • research_add_convergence_point — Register a convergence point
  • research_verify_convergence_point — Verify or reject convergence
  • research_record_contribution — Grade contributions (primary / auxiliary)
  • research_get_forest — Get full forest state and summary

Knowledge Tools

  • /ask-lit — Unified literature question answering
  • lit_search_local — Direct local literature search
  • code_qa — Code understanding based on evolution lineage

Experiment Tools

  • bench_adapt — Adapt code for a new benchmark or ablation
  • bench_run — Execute a benchmark in isolated worktree
  • bench_validate — Validate results against known SOTA

Visualization Tools

  • viz_generate — Generate analysis charts
  • viz_highlight — Highlight key data points
  • viz_polish — Polish charts for publication

Read the full file on GitHub · 144 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 2d ago First seen · 144 lines · 0 tokens per session scan A b4863167c82e

Subscribe to this mod's changes

research_agent is an agent published in the GitHub repository DataLab-atom/EvoAny (37 stars, last pushed 4mo ago), licensed Apache-2.0. It costs nothing until one of its globs matches a file; then it loads 1,200 tokens. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.