agentleFS
matthewdonsemail-labmatthewdonsemail-lab/log/.claude/skills/log-facebook-marketplace-listings/SKILL.md

log-facebook-marketplace-listings

Skill63 starsChanged 3 days agoMITView on GitHub

From accounts to live Marketplace listings. A log play (docs/useCases/facebook/marketplaceListings.md): How a local service business gets listings in front of people in every town it serves: take aged Facebook accounts, turn each one into a believable local profile, and post a listing from each one. This page walks one account all the way through, shows which part of log-cli does each step, and what decides that the step worked.

The file

We split this skill into the rules it gives an agent

5 rules. About 0 are common, 4 are distinctive, and 1 is 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.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Taken down: when needed: the delete answer names the listing id, and the row becomes REMOVED (rows are never deleted).

    line 40
  2. Very specific:

    Every Facebook command starts with the platform (log facebook …), and each has an MCP tool.

    line 46
  3. Distinctive:

    The proxy first: No browser opens without a proxy in the account's home country (rule 35).

    line 70
  4. Distinctive:

    Recorded: Every step is recorded, and Facebook's own answers decide what happened, never the screen (rule 73).

    line 73
  5. Distinctive:

    When a step lands somewhere new, its internal API calls are read, SubQ is asked what they mean, and the new state gets a name (rule one at …

    line 89

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill 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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

From accounts to live Marketplace listings

This is the log play facebook/marketplaceListings. Before anything else, call the log MCP tool getGuide({ useCase: "facebook/marketplaceListings" }): it returns this page and makes it the session's play, so every log tool's answer says which step you are on and a send with no play loaded is refused (rules 157, 158). Then follow the steps in order.

Draft every message from that person's own thread and matt's real replies (voice with no text, search), never from memory, and never send two that read alike (sameAsSent).

How a local service business gets listings in front of people in every town it serves: take aged Facebook accounts, turn each one into a believable local profile, and post a listing from each one. This page walks one account all the way through, shows which part of log-cli does each step, and what decides that the step worked.

How the owner asks for it: "post a marketplace listing in <town>", "get listings up on facebook marketplace in every town", "set up the facebook accounts and publish the listings", "take the marketplace listing down".

The idea in one line

One account per town. Each one gets a proxy in that town, its own browser profile, a photo, a cover, a name and a location, a day of normal browsing, and then one listing, all checked against Facebook's own answers.

flowchart LR
  buy["1. import the account<br/>Twenty socialAccounts"] --> proxy["2. a proxy near<br/>the home town"]
  proxy --> device["3. one AdsPower<br/>profile each"]
  device --> host["4. the host opens the browser<br/>once a day"]
  host --> setup["5. sign in, check,<br/>photo, name, location"]
  setup --> warm["6. warm-up"]
  warm --> publish["7. publish the listing"]
  publish --> check["8. check it every 12 min<br/>clicks into Twenty"]

One account, end to end

  • Imported into Twenty with its home town, cookies and two-step secret.
  • Proxy: a residential proxy near the home town, checked for country and distance before it's saved.
  • Location: typing "City, ST" can be offered as a different place entirely (a town of the same name abroad), and a bare zip can be offered as another country first. So locations are always typed with the state written out, "City, State" (spec).
  • Listing: the service, a price and photos. Facebook's create answer returns the listing id, under review at first, and Twenty marks it UNDER_REVIEW, then live.
  • Taken down when needed: the delete answer names the listing id, and the row becomes REMOVED (rows are never deleted).

Count the cost per account (the account, its proxy traffic, the browser profile) against what one lead is worth before scaling to many towns.

The steps and the commands

Every Facebook command starts with the platform (log facebook …), and each has an MCP tool. Anything that changes something on Facebook is its own …Live tool and needs confirmLive (rule 8).

#StepMCP toolWhat decides itCommand
1Import–the row in Twenty socialAccountsscripts/twenty/importVendorAccounts.mjs
2ProxyproxyAssign, proxyCheckthe exit IP is in the home country and within 100 km of the home townlog proxy assign --account <a> --vendor <vendor>, log proxy check
3Browser profileadsPowerSync (a dry run), then adsPowerSyncLive { confirmLive: true }one AdsPower profile per account, linked in adsPowerProfileIdlog adsPower sync --apply
4Ready?facebookPreflighthours, proxy, kernel, opens left, photo on disk, before anything openslog facebook preflight
5Open the dayfacebookHostStartLivethe browser's own timezone, language and exit IP match the accountlog facebook host start --account <a> --live
6Sign infacebookConnectFacebook's c_user and xs cookieslog facebook connect --account <a> (the host does it)
7Account checkfacebookCheck/marketplace/status says selling is openlog facebook check --account <a>
8Photo, cover, namefacebookProfileSetupLivethe photo and name saves answer without an errorlog facebook profile setup --account <a> --live
9LocationinspectDoLivethe location save's answer names the townlog inspect do setLocation "<City>, <State>" --live (the host does it)
10Warm-upwarmup, warmupLivethe five browsing activities each see their own answer; then 24 h before it's readylog facebook warmup stages run, log facebook warmup reel
11CopyfacebookListingCreate, facebookListingLintan ad on more than one listing in a city: at least 3 versions and one per listing live there, every two titles at least half different (the city not counted); campaign fill refuses too few (notEnoughCopyVersions)log facebook listing create, listing lint
12PublishfacebookFlowMarketplaceListingPublishLivethe create answer returns the listing idlog flow run marketplaceListingPublish --account <a> --live
13Many townsfacebookCampaignFill, facebookCampaignRunLiveone copy per account, a lost account's copy handed to another in the same arealog facebook campaign fill, run, heal
14WatchfacebookCheckeach listing's state and clicks from the Selling page, into agencyListingsthe host, every 12 minutes, or log facebook check
15Take downfacebookListingRemoveLive, facebookFlowMarketplaceListingDeleteLivethe delete answer names the listinglog facebook listing remove <id> --live

What keeps an account safe

  • One browser a day. The host opens each account's browser once per local day and keeps it open; every action is a tab in it (rule 35). On top of that, an account's browser opens at most 8 times a day, 10 minutes apart (rule 35).
  • The proxy first. No browser opens without a proxy in the account's home country (rule 35).
  • Local hours. Nothing runs outside 08:00–22:00 in the account's home town unless the owner asks (rule 35).
  • The same device every time. Each account keeps one fingerprint (rule 35).
  • Recorded. Every step is recorded, and Facebook's own answers decide what happened, never the screen (rule 73).

What log-cli does underneath

flowchart LR
  cmd["log facebook … / MCP tool"] --> gate["gates: proxy, hours,<br/>opens left, self-check"]
  gate --> browser["the account's AdsPower browser<br/>(one a day, one tab per action)"]
  browser --> fb["Facebook"]
  fb --> answer["Facebook's own answer<br/>read from the recording"]
  answer --> machine["the account's lifecycle machine<br/>onboarding lane, then host lane"]
  answer --> twenty["Twenty: account, listing, clicks"]
  • Each account runs one lifecycle machine (src/platformKit/lifecycle): proxying, opening, signingIn, checkingAccount, settingProfile, settingName, settingLocation, warming, attached, then watching and checking.
  • Flows are recorded once with log inspect and then replayed on any account (src/browser/flow/definitions/*.json, docs in docs/platforms/facebook/flows/).
  • When a step lands somewhere new, its internal API calls are read, SubQ is asked what they mean, and the new state gets a name (rule one at the top of AGENTS.md).

When accounts are lost

Not every account makes it. Some are lost before or after their first listing: a video selfie, a wrong password, or an identity check on a live listing. Their listings go to ACCOUNT_LOST, and a campaign hands each lost copy to another account in the same area (log facebook listing sweep). A checkpointed account is deactivated and gives its browser profile back.

Next

  • #41 reads buyers' messages into Twenty and replies to them (log facebook conversations keeps what is recorded today).
  • #42 publishes as a tab of the host's browser on every path (the recorded flow does already; the older log facebook publish opens its own browser).

Ways to reuse it

The pipeline doesn't care what the listing sells. Anything that wants a local profile in many towns uses the same parts: the proxy per town, the profile, the warm-up, the recorded publish flow, and the 12-minute check. A new kind of listing is a new copy set and photos (skills/listingCopy, skills/listingCreative), not new code.

Sources here: docs/lifecycle.md, docs/accountsAndListings.md, docs/proxies.md, docs/adsPower.md, skills/host/SKILL.md. What each step must do: the specs for proxies, the location, copy versions, the warm-up and publish and check.

Download now (ZIP)

Does 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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.