Research Desk: five rules for research you will act on, and a five-check script (free sample)
by PaperChaseWebb
From the Research bundle: five rules for research that feeds a decision (an owner and an auditor who did not do the research, a decision named for every question, one of five verdicts per claim, a date on every source, a primary source behind every CONFIRMED), and a read-only script that checks one small file for them.
- Price
- Free
- Difficulty
- Beginner
- Setup
- 5 min
- Format
- .md file
research-desk-five-rules-for-research-you-will-act-on-free-sample.md · 7 KB · direct download, no signup
What you end up with
In a few minutes you can see which of your agent's research answers are not ready to act on yet.
What you get
- Five research rules
- research_mini.py: five read-only checks on one research-lite.json
- One prompt for all four harnesses
- Worked example: 5 findings, then 0
Who it is for
Anyone who uses an agent to research before deciding something and wants a quick check of the basics first.
Works with
Solid chips: the agent is named in this listing's own tools list. Dashed chip: Plain Markdown, so it also pastes into other agents such as Grok Bot, Hermes, OpenClaw or Muse. We list those as compatible by format only, not as tested.
Tools needed
Cost to run: Free.
Time to result: About five minutes to a first result
Sample prompts, free to use
Read research-lite.json for the question we are working on. Run research_mini.py on it and show me the output. For each finding, tell me the smallest fix and what you need from me. Do not change verdicts or fill in dates you did not see yourself, and do not name yourself as the auditor.
The rules and the script are in the file.
Free preview of the file
# Research Desk: five rules for research you will act on, and a five-check script (free sample)
From the PaperChaseWebb Research bundle. Version 1.0, 2026-10-29. Plain markdown and one Python 3 script, standard library only. It reads one file, writes nothing and opens no network connection. Works with Grok Bot, Claude Code, Codex and Hermes through the prompt below.
## Install
Save this file as research-desk-sample.md, then pull the script out of it:
~~~sh
python3 -c "import re;t=open('research-desk-sample.md').read();open('research_mini.py','w').write(re.search(r'~~~python\n(.*?)~~~',t,re.S).group(1))"
~~~
Then write research-lite.json for your own question, or copy the worked example below, and run `python3 research_mini.py research-lite.json`. Exit 0 clean, 1 findings, 2 missing input.
## Five rules
1. Someone owns the decision, and someone who did not do the research reads the claims before anyone acts.
2. Every question says which decision it informs. A question that feeds no decision is not research yet.
3. Every claim gets one of five verdicts: CONFIRMED, CATEGORY ONLY, UNVERIFIED, CONTRADICTED or NOT CHECKABLE. "Probably true" is not one of them.
4. Every source has the date you opened it. A source with no date cannot be checked again later.
5. CONFIRMED needs a primary source: the maker's own docs, a filing, a dataset, a transcript. A roundup or a summary can lead you there but cannot confirm on its own.
## The five checks
| Check | What it looks for |
|---|Test record
- Date
- Oct 29, 2026
- What was run
- research_mini.py on the worked example, a fixed file and a missing file.
- Result
- 5 findings, then 0; exit 2 on a missing file.
- Not run
- The prompt inside a harness.
Copied from this item's dated test notes. What was not run stays in.
Worth knowing
- The sample checks one file against five rules. The full Research Desk, handoffs, shared registers, audits, approvals tied to the report hash and the fifteen checks are in the full bundle.
- On 2026-10-29 research_mini.py gave 5 findings on the worked example, then 0 on the fixed file, and exited 2 on a missing file. The prompt was not run inside a harness.
Questions
Does it send anything anywhere?
No. It reads one file and opens no network connection.
What is in the full bundle?
The Research Desk with fifteen checks and owner approval tied to each report's hash, plus Problem Mining Loop, Research Contract, Check the Claim and Compile Your Research in full. $59, against $101 for the four bought separately.
Details
Version 1.0 · Updated Oct 29, 2026 · Markdown version of this page
License: Personal and commercial use. Download terms
Related
Research bundle: from first complaint to a decision you can defend, on one desk a script can check
Problem Mining Loop, Research Contract, Check the Claim and Compile Your Research in full, tied into one loop by a new Research Desk: every question names the decision it feeds, moves one stage at a time only when that stage's own checker exits 0, and draws on one shared register of sources with snapshots and claims with five verdicts. Someone who did no research audits before anything is compiled, fifteen read-only checks name every gap, and the owner's yes is typed at their own terminal and tied to the sha256 of the exact report.
Research Contract for Agents: claims, sources and dates your agent must keep
A contract your agent follows on every research run, kept as plain files: each claim with its source, date and confidence, primary sources before secondary, quotes word for word with a location, contradictions listed, unknown allowed, and no citation to anything it did not open. A fifteen-check script reads the folder and names every break in a plain sentence, and nothing is published or acted on until the owner approves the exact report by its sha256.
Check the Claim (expanded edition): a method, prompts and a skill for checking AI claims
Make your agent check income, tool, popularity, quote, news and offer claims before they reach a brief, a pitch, a purchase or a post: label how much was read, run the creator's own arithmetic, trace sources back to first hand, and give each claim one of five verdicts on a card the owner approves. v1.1 adds an evidence level label (full, caption, summary or wall), four new arithmetic tests, a section on tracing a source back to first hand, screens for offers, quotes and slogans, four more red flags, a sixth worked example, a claim folder with a fourteen-check read-only checker and an owner approval helper, agent files for Grok Bot, Claude Code, Codex and Hermes, and twelve new prompts (26 in all). Price unchanged.