Skip to content
Agentmart
Submit a tool

research

Research across preprint, formal, web, and practitioner axes.

Skill by marcuskrogh

IndexedMIT★ 1
Write a review

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/marcuskrogh-research

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

Research across preprint, formal, web, and practitioner axes. Produces RESEARCH.md finding docs on the delivery branch (no separate PR). Use for surveys, state of the art, or a wide evidence pass.

Use cases: research

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/research

What's inside

Skill name
research
Description
Research across preprint, formal, web, and practitioner axes. Produces RESEARCH.md finding docs on the delivery branch (no separate PR). Use for surveys, state of the art, or a wide evidence pass.
License
MIT
Instructions
113 lines in SKILL.md · view on GitHub

Agent Skills spec issues

  • files: referenced but missing: scripts/arxiv_research.py

Files (2)

  • SKILL.md
  • reference.md

SKILL.md (first lines)

# Research

Applies [CONCEPT_RESEARCH](../concepts/CONCEPT_RESEARCH.md) and
[CONCEPT_SKILL](../concepts/CONCEPT_SKILL.md) as a **multi-axis**
investigation. Produces **finding docs**. A workflow may apply them as
supportive context. They do not settle product decisions.

**On invoke:** read [CONCEPT_RESEARCH](../concepts/CONCEPT_RESEARCH.md),
[CONCEPT_SKILL](../concepts/CONCEPT_SKILL.md),
[reference.md](reference.md), and [../workflow/SKILL.md](../workflow/SKILL.md).
When spawning axis workers, also read
[CONCEPT_DELEGATION](../concepts/CONCEPT_DELEGATION.md).

## Extensions

| Slot | This skill |
|------|------------|
| **Research axes** | Preprints (arXiv), Formal written, Web discovery, Informal / practitioner — all by default |
| **Retrieval path** | See table; arXiv details in [reference.md](reference.md) |
| **Artifact** | `RESEARCH.md` (path from WORKSPACE) |
| **Citation rules** | Every claim → retrieved evidence; durable ID/URL + **axis** label |
| **Pipeline continuity** | Commit `RESEARCH.md` onto the delivery branch; never open a PR |
| **Model routing** | CONCEPT_DELEGATION when axes are workers; synthesis stays on manager |

| Axis | Primary retrieval |
|------|-------------------|
| **Preprints (arXiv)** | `scripts/arxiv_research.py` first ([reference.md](reference.md)); curl/WebFetch API fallback only if Python unavailable |
| **Formal written** | WebSearch → WebFetch durable pages (DOI, publisher, IETF, ISO, W3C, …) |
| **Web discovery** | 2–4 complementary WebSearch queries; WebFetch primary pages |
| **Informal / practitioner** | WebSearch + WebFetch; **always label informal** |

## Steps

Follow CONCEPT_RESEARCH flow. Specialisations:

1. **Scope and plan axes** — Default to all four axes; when the user narrows the pass, record the included axes and reason. Write queries/targets into Search strategy. Done when every included axis has a retrieval plan and every skipped axis has a reason.
2. **Retrieve and synthesize** — Run axes in parallel when tools allow; otherwise Preprints → Formal → Web → Informal. Prefer one multi-`-q` arXiv search, dedupe by DOI / arXiv ID / canonical URL, and prefer multi-axis hits. Done when the CONCEPT_RESEARCH completion bars hold for retrieval, triage, deep read, and synthesis.
3. **Persist** — Write `RESEARCH.md` onto the delivery branch (no PR); when linked, apply the research tracker row. Record outcome `ready`. Apply the workflow transition. Done when claims are traceable and the artifact is on the delivery head with Task and mirrors agreed.

## Artifact

```markdown
# Research brief: <topic>

## Question
…

## Axes covered
| Axis | Status | Notes |
|------|--------|-------|
| Preprints (arXiv) | covered / skipped | … |
| Formal written | covered / skipped | … |
| Web discovery | covered / skipped | … |
| Informal / practitioner | covered / skipped | … |

## Search strategy
… (queries / targets per axis)

## Executive summary
… (what the sources say — not what we will build)

## Key sources
… (mixed axes; title, axis, ID/URL, one-line relevance)

## Themes and trends
… (agreements / disagreements across axes)

## Gaps and limitations
…

## Recommended reading order
… (sources to study — not a product plan)

## Role
Supportive finding doc. Not a product plan.

## Sources
… (full citations + axis)

## Tracker
- Task: <KEY> (if linked)
- Artifact: RESEARCH.md
- Branch: <delivery-branch>
- PR: — (research never opens a PR)
```

## Pipeline continuity

When a Task key is given/inferred:

Follow [finding-docs continuity](../workflow/delivery.md#rules) and the
[research tracker row](../workflow/tracker-sync.md#matrix). Commit
`RESEARCH.md` onto the **delivery** Task’s branch (create the branch if needed;
reuse the downstream delivery head when this is a supportive-only route Task).
**Never open a PR.** Leave delivery-Task status unchanged; mark a supportive-only
route Task **Done** at handoff. Update ROADMAP notes / PLAN Inputs when present.
Record the artifact, short summary, branch, and enabled mirror;
external artifacts are pushed into the Task without starting a branch.

## Inputs

| Input | When present | When absent |
|-------|----------------|-------------|
| Question | Apply it | Ask once what to investigate |
| Task, `ROADMAP.md`, or definition spec | Use as the question boundary | Investigate the question the user named |

## Output

`RESEARCH.md` — citable finding doc. Outcome: `ready`.

This skill does not name a successor. Apply the workflow transition before the turn ends. Standalone work still writes `RESEARCH.md`; with no bound workflow, **Next** is none.

From marcuskrogh/skills, MIT.

More from marcuskrogh/skills

Signals

Stars
1
Forks
1
Open issues
0
Last commit (this folder)
Sep 27, 2026 (1 month ago)

From GitHub (fetched Oct 24, 2026).

Details

License
MIT
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from marcuskrogh

Cite this page

research skill. Agent Mart. https://agentmart.dev/skills/marcuskrogh-research (last updated 2026-10-24)