research

A deep-research workflow that investigates a question across multiple sources, extracts cited claims, checks them against competing analyses, and produces a synthesis.

In plain words
What is it for?
Use it for multi-source research questions that need a fact-checked report, especially when the question has enough detail to investigate.
Why use it?
It reduces the manual work of searching, reading, comparing, and fact-checking many documents. If its local research tools are unavailable, it uses the built-in web-search path.

Skill for Claude CodeCodex

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 skills/robinslange/learning-loop/research
Any agent
npx skills add robinslange/learning-loop --skill research
Clone the repo
git clone --depth 1 https://github.com/robinslange/learning-loop

Made for: Claude Code, Codex.

Per session 66 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,358 The whole file, excluding the scripts and references it only reads on demand.
Security scan B 2 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 $0.00066 $0.01358
Opus 5 $0.00033 $0.00679
Sonnet 5 $0.00013 $0.00272
Haiku 4.5 $0.00007 $0.00136

Measured 2d ago against content hash fad389596d1b, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade B, and why

research scanned grade B with 2 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.

The scan reads SKILL.md. This mod also ships 1 executable file (workflow.js), listed below but not scanned — reading those needs a real analyzer, not pattern matching.

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.

Sends data to an external URLmediumData exfiltration

A POST to an outside endpoint may be telemetry or may be exfiltration; either way the mod talks to somewhere, and you should know where.

`curl -s localhost:11434/api/generate -d '{"model":"gemma3:12b","prompt":"hi","stream":false,"keep_alive":"30m"}' >/dev/null`

Makes network callslowCapability

Not a fault in itself. Listed so you know the mod talks to something, and to what.

`curl -s localhost:11434/api/generate -d '{"model":"gemma3:12b","prompt":"hi","stream":false,"keep_alive":"30m"}' >/dev/null`
plugin/skills/research/SKILL.md · 99 lines

How it starts

The opening of the file, as written. The whole thing — 99 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Research: librarian-offloaded deep research

Overview

Same shape as the built-in /deep-research — Scope → Search → Fetch → Extract → 3-vote adversarial Verify → Synthesize — but the middle three phases run on the local librarian (via bin/source-gateway.mjs research: Brave search + fetch + local Gemma claim extraction). Roughly 15 source documents are distilled to one-line cited claims before anything reaches Claude. Scope, Verify, and Synthesize stay on Claude — Verify is the step most likely to expose a small model's reasoning gap, and it runs over cheap one-line claims, not prose.

If the librarian can't run (Ollama down, no Brave key, sub-tier model, or it returns zero claims) the workflow falls back to the Claude-native WebSearch path so the command never hard-breaks.

When to use

  • /learning-loop:research "<question>" — a deep, multi-source, fact-checked report when you have a research-capable local model (12b+; chosen at /init).
  • If the question is underspecified (e.g. "what car to buy" with no budget/use-case/region), ask 2-3 clarifying questions first, then weave the answers into the question you pass.

Prerequisites (the workflow checks these, but know them)

  • Ollama running locally with a research-capable model resident (gemma3:12b default; the e2b tier is triage-only and will trip the capability gate).
  • Brave API key in the macOS Keychain (service="brave-search-api-key", account=$USER) — same key the Brave MCP uses.
  • The model and keep_alive come from librarian.* config (set at /init). Cold start: the first 12b call adds ~10-40s; keep_alive (default 30m) keeps it warm after. To avoid a cold first call, warm it first: curl -s localhost:11434/api/generate -d '{"model":"gemma3:12b","prompt":"hi","stream":false,"keep_alive":"30m"}' >/dev/null

How to run

First, resolve the plugin root${CLAUDE_PLUGIN_ROOT} is set in your main session but is NOT exported into Workflow subagent shells, so the workflow must be handed a concrete absolute path. In a Bash block, run:

Read the full file on GitHub · 99 lines

Files

What ships with it

1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.

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 · 99 lines · 66 tokens per session scan B fad389596d1b

Subscribe to this mod's changes

research is a skill published in the GitHub repository robinslange/learning-loop (11 stars, last pushed 11d ago), licensed Apache-2.0. It adds 66 tokens to every session and 1,358 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it B with 2 findings (sends data to an external url, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.

Related

Other skills, from other repositories

systematic-debugging

Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes.

obra/superpowers · 21 tokens

next-cache-components-adoption

Turn on Cache Components in a Next.js app and resolve the blocking routes it surfaces. Use when the user wants to enable, adopt, or migrate to Cache Components, flip the cacheComponents flag, work through a flood of blocking-prerender / instant validation errors, run the cache-components-instant-false codemod, or…

vercel/next.js · 95 tokens

babysit-pr

Babysit a GitHub pull request after creation by continuously polling review comments, CI checks/workflow runs, and mergeability state until the PR is merged/closed or user help is required. Diagnose failures, retry likely flaky failures up to 3 times, auto-fix/push branch-related issues when appropriate, and keep…

openai/codex · 114 tokens

imagegen

Generate or edit raster images when the task benefits from AI-created bitmap visuals such as photos, illustrations, textures, sprites, mockups, or transparent-background cutouts. Use when Codex should create a brand-new image, transform an existing image, or derive visual variants from references, and the output…

openai/codex · 113 tokens

cpu-profile-analysis

Analyze V8/Chrome CPU profiles (.cpuprofile) and DevTools trace files (Trace-.json). Use when: profiling performance, investigating slow functions, comparing code paths, finding bottlenecks, analyzing timeToRequest, understanding call trees from sampling profiler data, analyzing layout/paint/rendering, investigating…

microsoft/vscode · 71 tokens

next-cache-components-optimizer

Drive a Next.js route to instant navigation by setting up an agentic loop, under Cache Components / PPR, on initial load (hard navigation) and client-side navigation (soft navigation). Encode the goal as a failing @next/playwright instant() e2e and work it to green, one verified route at a time; the shipped test then…

vercel/next.js · 170 tokens