# Research Brief

> Use when gathering evidence before a decision, summarizing repo or external facts, checking whether context supports a proposed path, or preparing a brief before specs, plans, brainstorming, or implementation.

- Skill: `hassan-mohiddin/research-brief` (Agent Skill)
- Install (CLI): `npx skillmds@latest add hassan-mohiddin/research-brief`
- Raw SKILL.md: https://api.skillmd.com/api/skills/hassan-mohiddin/research-brief/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Research & Search
- Author: hassan-mohiddin (https://skillmd.com/u/hassan-mohiddin)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/hassan-mohiddin/research-brief

---


# Research Brief

Gather evidence before deciding.

## Target Gate

Before researching, identify the exact target: artifact, URL, repo, product, paper, package, model, organization, or source.

Use provided context first: links, screenshots, attached files, browser or page context, repo paths, issue numbers, and named artifacts.

If multiple plausible targets would change the research path, stop and ask for the source. Do not choose by search ranking, source richness, name similarity, or an "I assume..." caveat.

Research is not approval to follow the latest request.

If the request is biased, stale, or asks you to skip relevant sources, inspect the evidence anyway.

Do not edit implementation files. Do not create a durable brief unless asked.

## Evidence First

Start with the smallest sources that can answer the question:

- Code.
- Tests.
- Docs, specs, policies, ADRs.
- Logs, issues, traces, or prior reports.
- External sources when current facts matter.

Prefer live evidence over summaries, handoffs, or the user's framing.

## Brief Shape

Keep the brief short:

```text
Question: ...
Evidence: ...
Conflict: ...
Unknowns: ...
Decision needed: ...
Recommendation: ...
```

Omit sections that do not apply.

## Stop Conditions

Stop and surface the decision when evidence reveals:

- Ambiguous research target or source identity.
- Source-of-truth conflict.
- Missing product, scope, domain, compatibility, or public API choice.
- Security, privacy, billing, data-loss, permissions, or migration behavior.
- Enough uncertainty that more research would not pick the path.

Ask only after checking discoverable evidence.

## Handoff

If the brief will guide later work, link to evidence instead of copying volatile repo inventory.

If the next step is collaborative direction-setting, route to `grill-context`.

If the next step is durable requirements, route to `write-spec`.