Compile Your Research: a cited, checked second brain your agents keep
by PaperChaseWebb
Your agent compiles saved sources into linked markdown pages where every fact cites its file, conflicts get their own page, and nothing is the record until a person ratifies it. Four folders, ten rules, seven page types, ten prompts, a health check script that needs no AI, and a seeded test vault with nine planted problems.
- Price
- $24
- Difficulty
- Intermediate
- Setup
- 30 min
- Format
- .md file
compile-your-research-agent-second-brain-with-source-logs.md · 37 KB · instant download after checkout
Want this built for you instead? We can build it for you.
What you end up with
A research library you can ask questions of and get answers that cite the file they came from, with conflicts held open for you to decide and a script that tells you when something has drifted.
What you get
- 17 sections in one .md file
- A four-folder layout with a scope list so private material stays out
- Ten compile rules and a five-step approval flow: admit, propose, check, ratify, resolve
- Seven page types, including contradiction pages and a reading queue tied to what you build
- 10 copy-paste prompts
- wiki_check.py: a read-only health check for uncited lines, missing and unindexed sources, orphans, stale pages, open conflicts, status mistakes and out-of-scope files
- A seeded test vault with nine expected findings
- A weekly routine and harness notes for Grok Bot, Claude Code, Codex and Hermes
- A failure table, an illustrative worked example and a note on where it stops scaling
- A competitor teardown and credited sources
Who it is for
People who save more research than they can reread and want an agent to keep it compiled, cited and honest about conflicts, without a forty-template system they will stop using.
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: The pack is the only purchase. Uses the agent harness you already have and Python 3.
Time to result: About 30 minutes to set up and test on the seeded vault, then one compile per source and a weekly check
How it works
Set up
Make the four folders, paste the rules into the file your harness reads first, and save wiki_check.py.
Expected: An empty vault the agent can read and propose into.
Test
Build the seeded vault from section 10 and run the script, then Prompt 7 and Prompt 4.
Expected: Nine findings, and an answer that says the two prices disagree instead of picking one.
Scope
Run Prompt 1 on your own raw folder and write scope.txt.
Expected: A named list of files the wiki may compile, with email and private material left out.
Compile
Run Prompt 2 on each source, oldest first. Check proposals with the script.
Expected: Draft pages in proposed/, every Record line cited.
Ratify and ask
Move good proposals into wiki/, sign them, then ask questions with Prompt 4.
Expected: Cited answers, and contradictions waiting for your resolution.
Sample prompts, free to use
Compile raw/<file>, which is in scope.txt. Decide which wiki pages it touches and which new pages it needs. Write every new or changed page to proposed/ with status: draft and blank ratified_by. Each Record line ends with (src: raw/<file>). If this file disagrees with an existing page, write a contradiction page instead of choosing. Finish with pages created, pages changed, contradictions opened, lines you could not cite.
Prompt 2 of 10. Runs after the compile rules in section 4.
Answer from wiki/ only: <question>. Start at wiki/index.md. Cite the pages you used and the raw files those lines cite. If two pages disagree, say so and link the contradiction page. If the wiki does not hold the answer, say "not in the wiki" and name the source that would settle it.
Prompt 4 of 10.
Free preview of the file
# Compile Your Research: a cited, checked second brain your agents keep A research wiki by PaperChaseWebb for solo builders, researchers, consultants and small teams who collect more than they can reread. You drop sources into a folder. Your agent (Grok Bot, Claude Code, Codex or Hermes) compiles them into linked markdown pages where every fact cites the file it came from, conflicts get their own page instead of being smoothed over, and nothing becomes the record until a person ratifies it. Includes the folder layout, the compile rules, seven page types, ten prompts, a health check script that runs without any AI, a seeded test vault with nine planted problems, and harness notes for four tools. Plain .md. Version 1.0, 2026-10-27. Works with Grok Bot, Claude Code, Codex and Hermes, and pastes into Claude or ChatGPT. Obsidian is optional. ## Contents 1. Who this is for, and what it does 2. The idea in one paragraph, and what it will not do 3. Setup: four folders and three files, about 30 minutes 4. The compile rules and the approval steps 5. The seven page types 6. The ten prompts 7. The health check script 8. The weekly routine 9. Harness notes: Grok Bot, Claude Code, Codex, Hermes 10. The seeded test vault 11. Failure table 12. Worked example (illustrative) 13. When it stops working, and what to do then 14. When to bring us in 15. What we looked at, and what we did differently 16. Sources 17. License, refunds and version <!-- preview --> ## 1. Who this is for, and what
Worth knowing
- It will not remember what you never put in raw/, and the agent can still miss a page. The citation rule is what lets you tell.
- Plain-file wikis work well into the low hundreds of pages; section 13 covers what to do past that.
- The worked example and the seeded vault are invented. Neither is a client.
- On 2026-10-27 the seeded vault was built and wiki_check.py reported the nine expected findings, then 0 after the listed fixes. The prompts were walked through against it as a design check. Run the seeded vault in your own harness first.
Questions
How is this different from asking a chatbot about my files?
A chatbot searches the pile from scratch every time and can quietly pick the wrong note. Here each source is compiled once into pages that build up, every fact cites its file, and conflicts sit on their own page until you decide.
Will the agent change my notes?
Never your sources: raw/ is append-only. The agent writes proposals to proposed/; you move them into the wiki and sign them. The health check script only reads.
Do I need Obsidian?
No. It is plain markdown in folders. Obsidian is a nice way to browse the links, and any editor works.
What about private material?
Every vault starts with a scope list. Email exports, legal, health, money, family and client material without consent stay out by default, and the script flags any raw file not on the list.
Which tools does it work with?
Grok Bot, Claude Code, Codex and Hermes, with harness notes for each. The rules and prompts also paste into Claude or ChatGPT.
Details
Version 1.0 · Updated Oct 27, 2026 · Markdown version of this page
License: Personal and commercial use inside your own business, including wikis you keep for your own clients. Do not resell, republish or share this file. Download terms
Related
Contradiction pages and a read-only wiki check (free sample)
From Compile Your Research: five rules that stop an agent from picking a winner when your notes disagree, the contradiction page shape, a prompt to open one, and a read-only health check prompt. Works on any folder of markdown notes.
Check the Claim (expanded edition): a method, prompts and a skill for checking AI claims
A Markdown pack that makes your agent check income, tool, popularity, quote and funding claims before they reach a brief, a purchase or a post: arithmetic first, primary sources second, one honest verdict per claim.
Cross-Harness Relay: Grok Bot, Claude Code, Codex and Hermes on one job
A bot template and teaching guide for running more than one agent harness on the same work. Six seat cards (front desk, planner, builder, reviewer, tester, night shift), a RELAY.md seat map, a relay packet for every handoff, four loop shapes, routine-or-judgment routing with a fallback when limits run tight, a preflight check, the front desk and night shift split with a proof footer, shared memory rules, eight relay drills, a rules block and 14 prompts.
Agent Skills Workshop: write, test and run SKILL.md skills for Hermes, Claude Code and Codex
A bot template and teaching guide for building your own agent skills instead of collecting other people's. A skill card template, a three-skill split for long jobs (reference lock, plan with timings, build and critique), a loop that turns your corrections into skill drafts, an intake check for skills you did not write, four action classes, a SKILLS.md register, an overnight run sheet, eight skill drills, a rules block and 14 prompts.
Agency AI Ops Workbook: the Install sprint, self-serve
A two-week workbook for owner-led agencies and small businesses: install one AI agent workflow yourself with the same steps as our $5,000 Install sprint. Discovery and mapping, picking the first workflow from your Self-Check score, an agent card with an owner and a tested pause, data access and a test set, an approval gate register, handoff packets, cost limits, tool vetting, a baseline and stop rule, a dry run, a go-live checklist and a 30-day review. Ten worksheets, 13 prompts with harness notes.
Agent Readiness Self-Check: score one workflow before you hand it to an AI agent
A free self-check for agencies and small businesses. Score one workflow across eight areas (workflow mapped, data access, approval gates, owner per agent, handoffs, cost tracking, security and vetting, measurement) with an evidence rule and three gate rules, read your band and what to do next, plus an interview prompt for Grok Bot, Claude Code, Codex or Hermes, a score sheet and a worked example.
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.