research-brief-lite
ASecurityFree check for a Markdown research brief. Finds a missing question, search budget, stop rule or findings section, a claim with no label ([V] verified, [I] inferred, [U] unverified, [X] not found) and a verified claim with no source. Use when an agent researches a market, a tool or a question, when the user asks for a research brief or a fact-checked summary, or before sharing research. Search words: research brief, fact check, source check, citations, search budget, stop rule, deep research.
- 0 votes
- 1 copy
- 4 views
- Added October 29, 2026
Works with
Security analysis
100/100Pro scans all 7 files and shows the line behind each finding
npx -y skills add MalikKhaledSaid/agent-skills --skill research-brief-lite --agent claude-codeAre you the author of research-brief-lite?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/malikkhaledsaid-research-brief-lite)---
name: research-brief-lite
description: Free check for a Markdown research brief. Finds a missing question, search budget, stop rule or findings section, a claim with no label ([V] verified, [I] inferred, [U] unverified, [X] not found) and a verified claim with no source. Use when an agent researches a market, a tool or a question, when the user asks for a research brief or a fact-checked summary, or before sharing research. Search words: research brief, fact check, source check, citations, search budget, stop rule, deep research.
license: MIT. LICENSE has complete terms.
compatibility: Designed for Claude Code; works with any agent that reads files. Needs python3 (3.8+). No network access.
metadata:
version: "1.0.1"
edition: "free"
---
# Research brief (lite)
You run research as a brief: one question, a budget and a stop rule first, then a label on every claim. The script checks the brief before it is shared.
## Step 1: Set up the brief with the user
Before any search, agree on one question, a search budget (for example "Searches: 8") and a stop rule (for example "Stop when 3 independent sources agree, or when the budget is spent"). Write them in the layout in `references/format.md`. Stop searching when the stop rule says so.
## Step 2: Label every claim
One claim per item under `## Findings`: `[V, source, date]` if you read it in the source, `[I]` if worked out, `[U]` if not checked, `[X]` if not found (say what you tried). `references/labels.md` explains each label. Never invent a source.
## Step 3: Check the brief
```bash
python3 <skill-folder>/scripts/brief_lint.py BRIEF.md
```
ERROR for a missing question, budget, stop rule or findings, and for a claim with no label. WARN for a verified claim with no source and for more than one question. `--json` prints JSON. Exit code 1 means errors.
## Limits
- A file with no brief sections is read as research notes: every labelled line is checked for a source.
- The check reads Markdown with level-2 headings. It cannot tell whether a source supports the claim.
- This edition does not count searches against the budget, check dates or listed sources, or compare before and after. The full edition does.
Full edition: research-brief $9: start a brief, log each search against the budget and stop when it is spent, checks for dates, listed sources and over-budget logs, a fix guide for each finding, a CLAUDE.md block, a worked example and a before-and-after check. https://buy.polar.sh/polar_cl_c3IMazGk7ocx6JpKwmUcGP8WPms58i1pAnAbz2QTI43
Files in this skill
- .claude-plugin/plugin.json
- LICENSE
- README.md
- SKILL.md
- references/format.md
- references/labels.md
- scripts/brief_lint.py
Attribution
Comments
Loading comments…
Newsletter
Get the weekly list of safe new skills, and the ones we pulled.