Skip to content
Agentmart
Submit a tool

research

Research that changes NOTHING.

Skill by stage2dev

IndexedMIT
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/stage2dev-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 that changes NOTHING. Use for "research only", "don't change anything", "study first", "compare with our code", "só pesquisa", "não altere nada", "estude primeiro". Ends in numbered questions.

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

Tools it pre-approves 4

From allowed-tools in SKILL.md: the agent may use these without asking you each time.

  • Bash(sh
  • "${CLAUDE_PLUGIN_ROOT}/scripts/py"
  • scripts/guarda.py
  • info)

What's inside

Skill name
research
Description
Research that changes NOTHING. Use for "research only", "don't change anything", "study first", "compare with our code", "só pesquisa", "não altere nada", "estude primeiro". Ends in numbered questions.
License
MIT
Allowed tools
Bash(sh, "${CLAUDE_PLUGIN_ROOT}/scripts/py", scripts/guarda.py, info)
Instructions
50 lines in SKILL.md · view on GitHub

Before installing

  • Pre-approves some shell commands

Files (1)

  • SKILL.md

SKILL.md (first lines)

# Research - read only

Context: !`sh "${CLAUDE_PLUGIN_ROOT}/scripts/py" scripts/guarda.py info`

Question: $ARGUMENTS

The JSON above is the context. Write to the user in its `language`, honoring `reply_style`. Use document names from `names`, never literal. Models/effort per role from `models`; if missing use the table defaults.

Nothing is changed: no code, database, external service or git (no commit, push, PR, stash). Write only in the scratchpad or the plans folder (`plans_dir`).

## Step 1 - Understand

- Ambiguous question: echo in 1 line what you will answer and go on. Do not stop to ask what can be assumed.
- Read the project profile (`profile_path`) if present: diagnostic sources, production, pitfalls.
- Define the base: `git fetch` and the hash of `origin/<main>`; always read code from it, not from the dirty tree.

## Step 2 - Dispatch (one message, background)

| Purpose | Agent | Default model | Effort |
|---|---|---|---|
| Where it is / how it works in our code | `tune-up:mapper` | haiku | medium |
| Official docs, market, competitor, pricing | `tune-up:researcher` | sonnet | medium |

- Simple question: 1 mapper is enough; 2-4 only for a wide topic. One theme each, numbered questions, "answer with file:line" or "URL + date".
- Paid tool (SEO, video, credit API): estimate cost first and save raw data to disk BEFORE summarizing.
- Never a model in `forbidden_subagent_models` (default `["fable"]`). Do not wait awake; do not read the agent's output file.

## Step 3 - Measure (yourself)

- Database and logs: READ ONLY (select, list logs). No insert, update, migration, deploy.
- A measured number beats a guess. Note the query and date.

## Step 4 - Consolidate

Write `<scratchpad>/research-<topic>.md` (or in the phase folder, if one exists; named by `names.research` there):
- 3-line answer on top.
- Findings with `file:line` (at the cited hash), measured numbers and source (URL + date).
- What was NOT verified.
- Options with cost and risk.

## Step 5 - Answer

1. The finding that changes everything, in 1-2 sentences.
2. Numbered questions, recommended first, answerable by "1 yes / 2 no".
3. Absolute path of the saved file.
4. Final sentence: "Nothing was changed." Only if true; if something was touched, say what.

Research mode ends when the owner says "let's build", "go ahead", "pode fazer" or equivalent. Then use `/tune-up:plan` or `/tune-up:part`.

From stage2dev/tune-up, MIT.

More from stage2dev/tune-up

Signals

Stars
0
Forks
0
Open issues
0
Last commit (this folder)
Oct 20, 2026 (8 days ago)

From GitHub (fetched Oct 24, 2026).

Details

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

More from stage2dev

Cite this page

research skill. Agent Mart. https://agentmart.dev/skills/stage2dev-research (last updated 2026-10-24)