Monitors 15 RSS feeds, deduplicates against a persistent article log, and writes new stories to Google Sheets via an API endpoint.

Rewrites articles using a 3-layer enrichment cascade, publishes to WordPress, then self-checks 4 mandatory SEO fields and auto-fixes any failures.

Generates branded social images from 6 rotating templates, then posts to Instagram, Facebook, and Lnk.Bio simultaneously. One post per run for natural feed spacing.

Generates a visual recap carousel — cover slide, 5 top stories, and call-to-action — from the week's published posts. Requires at least 3 posts to fire.

Scans ~964 followed artists on Spotify plus 5 indie-press outlets for releases in a 7-day forward window. Writes 10 ranked candidates to the review queue.

Generates 2 ranked editorial list ideas with carousel layouts, SEO targets, and article outlines. Uses traffic analytics insights to orient proposals toward what's working.

Audits 5–10 pages daily: title tags, meta descriptions, internal links. Fixes what it can automatically and flags what needs manual attention.

Closes the editorial week, analyzes traffic and search performance, and programs next week's content calendar. Writes a weekly plan and appends to cumulative strategic memory — the feedback loop that makes the whole system learn.

Daily competitive scan of 8 indie-music outlets. Cross-references Indietheka's coverage to identify gaps and writes the highest-value opportunities directly to WordPress as drafts — no approval gate.

Classifies posts by content type (Article / Review / NewsArticle / MusicEvent), generates JSON-LD structured data, and PATCHes WordPress meta. Day-1 deployment enriched 119 posts with zero failures.

Validates every published post against rules.json — Yoast compliance, featured image presence, internal links, schema validity, slug structure, tag taxonomy. Logs every run; the log is the source of truth for the Quality Monitor.

Audits the agent ecosystem itself. Scans Cowork transcripts, eval log, distribution log, MEMORY.md, and the agent repo for drift between spec and runtime. Detects error patterns not yet codified, proposes new rules — never writes, only proposes.

Pulls GA4 + Search Console for the closing week, generates the 11-component dashboard (bounce-by-page, keyword opportunity matrix, GSC↔GA4 anomalies, cohort retention, path analysis, schema CTR). Bloque A executes 4–6 metadata PATCHes per run autonomously.

Comprehensive site-wide audit across 8 dimensions (Technical, E-E-A-T, On-Page/SXO, Schema, CWV, AI/GEO, Sitemap, Images). Produces a scored Health Report routed to AGT-08, AGT-11, or Abe. Scorer, never fixer — every finding ends with an explicit Owner.