Analytics Collector

Analytics Collector

You measure what happened after publishing. Given the ids of posts that went out, you pull their engagement from X and LinkedIn and normalize each to one Metric. You are the analytic in trend -> remix -> produce -> post -> analytic -> repeat.

This is where the loop closes. The orchestrator pipes each Metric you emit into gf record-metric ; the connector recomputes that item's feedbackScore; and gf feedback turns those scores into a multiplier that lifts previously-successful themes to the top of the next trend scan. You produce the raw measurement that makes the factory learn.

Pure worker: reads one token from the environment, persists nothing, writes no manifest itself.

---

Run It

bun run scripts/collect.ts --platform x --ids 1799111,1799222
bun run scripts/collect.ts --platform linkedin --ids "urn:li:share:6844785523593134080"

A ContentItem (or an array of {platform, postId}) on stdin is an alternative to --ids — every derivative with a postId is collected:

bun run ../genfeed-connector/gf.ts get  \
  | bun run scripts/collect.ts --platform x

Requires Bun 1.1+. Zero dependencies — only Node built-ins and global fetch.

---

Platforms

X

• Endpoint: GET https://api.x.com/2/tweets?ids=...&tweet.fields=public_metrics,organic_metrics,non_public_metrics.
• public_metrics works with an app-only or user token; organic_metrics / non_public_metrics (impressions, link clicks) need a user-context token for the author. If the richer fields are denied, it automatically retries with public_metrics alone.
• Mapping: impressions with LinkedIn-Version + X-Restli-Protocol-Version: 2.0.0.
• Mapping: `likes metrics.json

2. record each metric — this recomputes feedbackScore on the item

jq -c '.metrics[]' metrics.json | while read -r m; do bun run ../genfeed-connector/gf.ts record-metric "$ITEM_ID" { "term": "ai agents", "multiplier": 0.74 }

`collect` measures, the connector remembers, `trend-scout` re-ranks. That three-skill cycle is the locked loop.

---

## Security

- Tokens (`X_BEARER_TOKEN`, `LINKEDIN_ACCESS_TOKEN`) are read from the environment, used in memory, and **never written** anywhere. Resolve them through the connector.
- All calls are read-only GETs against the platform analytics endpoints; this worker never posts, edits, or deletes.

## Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [genfeedai](https://github.com/genfeedai)
- **Source:** [genfeedai/skills](https://github.com/genfeedai/skills)
- **License:** MIT
- **Homepage:** https://skills.sh/genfeedai/skills

Install and usage instructions live in the source repository linked above.