agentleFS
graphed-incgraphed-inc/marketing-skills/skills/post-engagers/SKILL.md

post-engagers

Skill73 starsChanged 3 days agoMITView on GitHub

Deploy a hosted Post Engagers app on Graphed: paste a LinkedIn post, or track LinkedIn profiles daily, and get everyone who reacted or commented as a lead list. Each person gets a full profile and an email from a GetLeads → LeadMagic → Apollo waterfall, checked by MillionVerifier, in the project's Postgres, with a searchable table and CSV export. By default ask for a LinkedIn post or profile and queue the first list from the terminal; do not send them into the app to paste the URL. Use when the user wants a LinkedIn engager tool for their team, wants to track a profile's net-new engagers every day, or invokes /graphed:post-engagers.

  • Reads credentials
  • Installs packages

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

11 rules. About 0 are common, 5 are distinctive, and 6 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 6

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Tracking: add a LinkedIn profile. Every morning a cron job pulls that profile's new posts and runs each one through the same pipeline.

    line 14
  2. Very specific:

    If the user only wants a one-off list written into an existing scaffold project, use /graphed:enrich-linkedin-engagers instead.

    line 22
  3. Very specific:

    Use this skill's own directory.

    line 26
  4. Very specific:

    If graphed is not on PATH, stop and ask the user to install it (npm i -g @graphed-inc/cli) and run graphed login.

    line 43
  5. Very specific:

    Ask which Graphed account to deploy to (graphed --format json accounts list).

    line 45
  6. Distinctive:

    Read the tool contracts once.

    line 48
  7. Very specific:

    Never add public.paths to the manifest.

    line 80
  8. Distinctive:

    Do this from the terminal.

    line 94
  9. Very specific:

    Do not use the graphed exec output to move files or data.

    line 157
  10. Distinctive:

    Ask the user before you queue a post with more than about 2,000 engagements, set a cap above 2,000, or track more than one or two profiles.

    line 182
  11. Distinctive:

    Every tool call carries an idempotency key built from a random per-database id, the post id, the job, and the attempt.

    line 208

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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

Post Engagers

A small web app the user deploys to their own Graphed project. It turns LinkedIn engagement into a verified lead list:

  • One post: paste a post URL. The app pulls everyone who reacted or commented (replies included), keeps one row per person, opens each full profile, finds an email, and verifies it.
  • Tracking: add a LinkedIn profile. Every morning a cron job pulls that profile's new posts and runs each one through the same pipeline. "Net new" means people first seen in a time window (today, this week, this month). The app shows them per profile and in one list for everyone, and exports them as CSV.

The app sits behind the Graphed login, so the user's teammates on that account can open it. There is no pricing page: every scrape bills the project's Graphed credits.

If the user only wants a one-off list written into an existing scaffold project, use /graphed:enrich-linkedin-engagers instead. This skill is for a hosted tool with a UI and daily tracking.

Use this skill's own directory. When the skill loads, that directory is the "Base directory for this skill". Below, <skill-dir> means that path.

<skill-dir>/
  app/                         the deployable app (copied into the user's project folder on deploy)
    packages/core              schema, worker, email waterfall, tracker, JSON API
    packages/dashboard         Next.js UI at / and the /api catch-all
    cli.mjs                    queue, status, track, export (run with graphed exec)
    smoke.mjs                  read-only API smoke test inside the deployed image
    graphed.yaml               service, release migration, daily track cron
  scripts/deploy.sh            copy app/ into a folder, install, test, graphed deploy
  scripts/pull-csv.sh          export a list to project storage and download it

Before you start

  1. If graphed is not on PATH, stop and ask the user to install it (npm i -g @graphed-inc/cli) and run graphed login. Node 20+ must be on PATH.
  2. Ask which Graphed account to deploy to (graphed --format json accounts list). Ask for a project slug. The default is post-engagers. Check that the slug is free with graphed --format json projects list --account <account_id>.
  3. Read the tool contracts once. Do not swap in other actors.

    graphed tools catalog detail apify:apimaestro~linkedin-post-reactions
    graphed tools catalog detail apify:apimaestro~linkedin-post-comments-replies-engagements-scraper-no-cookies
    graphed tools catalog detail apify:apimaestro~linkedin-profile-posts
    graphed tools catalog detail apify:harvestapi~linkedin-profile-scraper
    graphed tools catalog detail getleads:enrich.from-linkedin
    graphed tools catalog detail leadmagic:people.email-finder
    graphed tools catalog detail apollo:people.bulk_match
    graphed tools catalog detail million-verifier:validate
  4. Tell the user what it costs (table below). Deploy only after they say yes.

1. Deploy

bash <skill-dir>/scripts/deploy.sh --account <account_id> [--slug post-engagers] [--name "Post Engagers"] [--dir apps/<slug>]

The script copies app/ into apps/<slug>/ in the current directory, runs npm install and npm test, and deploys. It prints the service URL. A --dry-run flag validates the manifest without deploying. The cloud build runs npm run db:migrate before the Next server starts. The page at / is the app. Health is GET /api/health.

  • Running it again is an update. Code is copied again from <skill-dir>/app/. The folder's .graphed.json and .git stay: graphed deploy pushes that repo, and deleting it causes a merge conflict on the next deploy. The cloud Postgres keeps every job, person, and tracked profile.
  • To change the app, edit <skill-dir>/app/ and deploy again. Edits made only in apps/<slug>/ are overwritten.
  • Never add public.paths to the manifest. The app has no auth of its own; the Graphed gateway is its login.

Prove the deploy before you hand it over:

graphed exec --project <slug> --account <account_id> -- node smoke.mjs

Its output ends with SMOKE ALL PASS. That command starts a second copy of the server with the worker off, calls the read endpoints, and writes nothing.

2. Set it up

Do this from the terminal. The page is how the user looks at the list afterward. Do not ask them to paste a URL and click Scrape engagers or Start tracking.

Ask your agent, in the top right of the app, copies a prompt. It explains what this app can build and tells you to queue the first list yourself. Follow that prompt, and the commands below.

Default: one post or one profile

Unless the user already gave a LinkedIn post URL or a profile URL, ask one question: a post to scrape once, or a profile to watch every morning. Then queue it.

The first scrape uses a profile cap of 25 unless they name a different cap. Tell them the cost before you queue. Plan on about 1 to 1.5 credits per engager before email lookups, then about 0.9 to 3.5 more for each person who reaches Apollo or LeadMagic. Do not raise the cap, queue a post with more than about 2,000 engagements, or track more than one profile until they agree.

graphed exec runs inside the deployed app. Each stdout line is prefixed with [runtime …] .

graphed exec --project <slug> --account <account_id> -- node cli.mjs add "<post url>" --cap 25
graphed exec --project <slug> --account <account_id> -- node cli.mjs track add https://www.linkedin.com/in/<user> --backfill 7 --cap 25

The web worker runs the scrape. Closing the terminal does not stop it. Adding a tracked profile starts its first check immediately. Poll until status shows the job done or failed:

graphed exec --project <slug> --account <account_id> -- node cli.mjs status

done is the first list. failed means show that job's error and stop. Then reply with the app URL, what you queued, and the counters: engagements, unique people, new people, profiles, emails found, verified.

3. Use it

In the app: New scrape (one post plus a profile cap), Tracking (profile URL, backfill days, profile cap; Check now, Pause), Everyone (lifetime list with net-new filters). Every view has Export CSV. Ask your agent stays in the top right.

From the terminal. The web worker does all the scraping, so closing the terminal never stops a job.

E="graphed exec --project <slug> --account <account_id> --"
$E node cli.mjs add "<post url>" --cap 1000
$E node cli.mjs track add https://www.linkedin.com/in/<user> --backfill 7 --cap 1000
$E node cli.mjs track list
$E node cli.mjs track check [tracked-id]
$E node cli.mjs track off <tracked-id>
$E node cli.mjs status [job-id]
bash <skill-dir>/scripts/pull-csv.sh <job-id|all> --project <slug> [--since 7] [--tracked <id>]
graphed jobs run track --project <slug> --account <account_id>

The last command runs the daily check now.

Accepted post URLs: linkedin.com/posts/<slug>-activity-<id>-…, feed/update/urn:li:activity:<id>, urn:li:ugcPost:<id>, urn:li:share:<id>, or the bare post id.

Do not use the graphed exec output to move files or data. CSVs go through project storage (pull-csv.sh does that). Confirm a write by reading it back with status or track list. For a one-off query, run it through tsx from the project root: node --import tsx -e 'import { pool } from "./packages/core/src/db/sql.ts"; …'.

Cost (1 credit = 1¢)

StepGraphed toolCredits
Tracked profile check (20 posts)apify:apimaestro~linkedin-profile-posts0.575 / post
Reactions, 100 per pageapify:apimaestro~linkedin-post-reactions0.575 / reaction
Comments and repliesapify:apimaestro~linkedin-post-comments-replies-engagements-scraper-no-cookies0.575 / comment
Profile, no emailapify:harvestapi~linkedin-profile-scraper0.414 / profile
Email 1: GetLeadsgetleads:enrich.from-linkedin, 25 per callabout 0
Email 2: LeadMagicleadmagic:people.email-finder3.45 / email found
Email 3: Apolloapollo:people.bulk_match, 10 per call0.92 / person matched
Verifymillion-verifier:validate0.046 / email

Each email provider runs only for people who still have no MillionVerifier ok email. Before email lookups, plan on about 1 to 1.5 credits per engager. Add 0.9 to 3.5 for each person who reaches Apollo or LeadMagic. The form's profile cap defaults to 1,000 per post. The first list you queue in Set it up uses 25 unless they name another cap. The cap limits the profile and email steps only. People profiled in the last 30 days are reused across posts at no cost (PROFILE_REFRESH_DAYS), and their email is not looked up again unless their company changed.

Ask the user before you queue a post with more than about 2,000 engagements, set a cap above 2,000, or track more than one or two profiles.

Waterfall order comes from the EMAIL_WATERFALL env. The default is getleads,leadmagic,apollo. getleads,apollo,leadmagic costs less per email. If you add harvest to the list, the profile scrape switches to its email-search mode (1.15 instead of 0.414 per profile). Any email that a profile already has is always verified first, with no lookup.

Setting these options: EMAIL_WATERFALL, PROFILE_REFRESH_DAYS, MAX_PROFILES_PER_JOB (web service) and TRACK_MIN_AGE_HOURS (web service and the track job) are read from the environment. To set one, add its name under services.web.env (and jobs.track.env when the job uses it) in <skill-dir>/app/graphed.yaml, run graphed secrets set <NAME> --project <slug> --account <account_id>, then deploy again. A listed name with no secret value fails the deploy, so add only the names you set.

MillionVerifier: ok counts as verified and ends the waterfall for that person. catch_all is kept as a fallback while later providers try. invalid and disposable are dropped. The "Verified email" filter and the verified counters are ok only.

How the worker runs

  • One job at a time. The job row stores which stage is done, so a redeploy resumes a job at its stage. Each finished page of reactions or comments is saved with its counts, so a restart does not add those people or credits again. The next worker reclaims a running job after 3 minutes with no heartbeat.
  • Every tool call carries an idempotency key built from a random per-database id, the post id, the job, and the attempt. A resumed stage replays runs that already finished instead of paying again. Profile lookups keep the original batch, and email lookups hash the full batch, so a restart cannot replay a result against a different set of people. Two deployments on one account never replay each other's runs.
  • A provider outage, timeout, or missing credits leaves those people unfinished. The job stops and can be retried. A confirmed "no email" does not.
  • Retry on a failed job starts fresh runs. Retry on a finished job refreshes profiles and emails for up to the profile cap, commenters first. People past that cap keep the profile and email they already have, including when another post shares them.
  • One row per person, keyed on the LinkedIn member id (ACoAA…). The reactions scraper returns only that id. The profile scraper takes it through profileIds and resolves the public /in/ URL, so reactors are enriched too. Company pages that react are skipped.
  • Tracking scrapes each post once, after it is 20 hours old (TRACK_MIN_AGE_HOURS), so most of its engagement is in. Posts by other authors in the feed are skipped. The first check of a new profile takes posts from the last backfill days. A post that was already scraped by hand is attached to the tracked profile, so its people show up in that profile's list. Two checks at once cannot queue the same post twice.

Tool quirks (handled in the code)

  • profileScraperMode must be exactly Profile details no email ($4 per 1k) or Profile details + email search ($10 per 1k). Other values return 400.
  • GetLeads often matches the person but returns no email, and its company data can be out of date. Its company size is used only when its company name matches the live LinkedIn one.
  • LeadMagic needs the employer's domain. It comes from the profile's website list (link-in-bio hosts are filtered out), then from GetLeads.
  • Apollo bulk_match takes camelCase details[], and matches[] lines up with it.
  • The comments scraper's last row is a summary object. Replies are in each comment's replies[].

Done

Reply with the app URL, the smoke test result, and, for a scrape, the job counters: engagements, unique people, new people, profiles, emails found, verified. If a stage failed, say which one and show the job's error.

About Graphed

Graphed is a marketing cloud. It includes the infrastructure for an agent to do marketing: a data warehouse, data pipelines, a Postgres database, cron jobs, runtime servers, a tool catalog, custom domains, and internal app tooling.

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.