reverse-engineer
AGENTS.md1 starsChanged 34 days agoMITView on GitHub
How to use this repository from any coding agent (Codex, Cursor, Claude Code and others). No install is needed. Read the skill file, then follow it. Eight skills in skills/<name>/SKILL.md and a starter app in template/. The sign-in happens in the browser. Never print, log or commit a token. Plain words, sentence case, short sentences. No em dashes. No hype. Every number has a source or a run id.
- Installs packages
Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.
Of the AGENTS.md files we read, the 31,542 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.
In this repository9 other files it gives its agents.
- llms.txtllms.txt
- bounded-probeSkill
- claims-auditSkill
- competitor-diffSkill
- evidence-researchSkill
- rebuild-on-loootSkill
- request-providerSkill
- reverse-engineerSkill
- site-teardownSkill
The file
We split this AGENTS.md into the rules it gives an agent
8 rules. About 0 are common, 4 are distinctive, and 4 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
Distinctive · estimate
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 23
Start with the free path: plain GETs of public pages, GET https://api.looot.ai/v1/catalog/overview?depth=jobs and GET …
- Distinctive: line 25
Before any paid looot call, quote it and show the user the number.
- Very specific: line 33
No MCP: install the CLI with npm install -g looot, or call the REST API at https://api.looot.ai with Authorization: Bearer $LOOOT_TOKEN.
- Very specific: line 40
Every run carries a stable idempotencyKey.
- Very specific: line 42
Read status, not only the HTTP code.
- Distinctive: line 49
Read public scripts as text.
- Distinctive: line 51
Do not copy the target's code, text, logos or screenshots.
- Distinctive: line 52
Do not publish, deploy or file anything without the user's yes.
≈ “Common” is an estimate: at least 1 in 5 of the 40,246 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
AGENTS.md
How to use this repository from any coding agent (Codex, Cursor, Claude Code and others). No install is needed. Read the skill file, then follow it.
What this repo is
Eight skills in skills/<name>/SKILL.md and a starter app in template/.
| Skill | Use it when the user wants to |
reverse-engineer | understand how a product, feature or go-to-market motion works from public evidence |
rebuild-on-looot | build the looot version of a product from a reverse engineering report |
request-provider | ask looot for a data source that the catalog does not have |
bounded-probe | test a looot job on a few inputs before a bulk run |
site-teardown | take apart a public website: pages, price, stack, ads, search |
evidence-research | get a company or market brief where every claim has a source |
competitor-diff | compare a competitor with its own earlier snapshot |
claims-audit | find the claims in a document that no source backs |
Procedure
- Open the
SKILL.mdthat matches the request. Follow its steps in order. - Start with the free path: plain GETs of public pages,
GET https://api.looot.ai/v1/catalog/overview?depth=jobsandGET https://api.looot.ai/v1/public-catalog?q=<words>. These need no token. - Tag every finding observed, inferred or unknown, with its source.
- Before any paid looot call, quote it and show the user the number. Ask for a yes above $0.10 or above 100 calls.
- For a full job, run the skills in this order:
reverse-engineer, thenrebuild-on-looot, thenrequest-providerfor anything missing.
Connect looot (optional, for paid jobs)
- Claude Code:
claude mcp add --transport http looot https://api.looot.ai/mcp - Codex:
codex mcp add looot --url https://api.looot.ai/mcp, thencodex mcp login looot - Cursor: add
{"mcpServers":{"looot":{"url":"https://api.looot.ai/mcp"}}}to~/.cursor/mcp.json - No MCP: install the CLI with
npm install -g looot, or call the REST API athttps://api.looot.aiwithAuthorization: Bearer $LOOOT_TOKEN.
The sign-in happens in the browser. Never print, log or commit a token.
looot call rules
- Quote first.
inspectis free and shows the exact input names and the maximum cost. There is no dry-run call. Inspect is the quote. It needs a sign-in. Without one, read the price from the public catalog. - Every run carries a stable
idempotencyKey. Reuse it to retry the same call. The same key with a different input returns409 idempotency_conflict, so use a new key for a new input. - A job id looks like
job:company.enrich. Run it withfallback: trueandmaxCostUsdset to twice the quoted price. - Read
status, not only the HTTP code.queued,running,pending_providerandreconciliation_pendingmean pending, not failed. A failed call costs nothing. - A 402 means the balance is low. Show the user the top-up link (minimum $5). Never pay for them.
- Wrong or empty answer: send
feedbackwith the run id and no personal data.
Hard limits
- Public information only. No logging in to accounts you do not own. No bypassing access controls. No scraping behind a login.
- Read public scripts as text. Never run untrusted code.
- Respect robots.txt and terms. Send nothing to the target.
- Do not copy the target's code, text, logos or screenshots. Do not imitate its brand.
- Do not publish, deploy or file anything without the user's yes.
Writing rules
Plain words, sentence case, short sentences. No em dashes. No hype. Every number has a source or a run id.
Check your work
npm run check # lints the skills and checks every job id against the live catalog
cd template && npm test && npm run typecheck && npm run lint && npm run buildDoes it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.