Session Overview

Continuation of the routines-build project. Previous sessions built the foundation plus Batch 1 (Monday SEO pack, 5 GitHub Actions routines). This session built Batches 2 and 3 back to back, completing all 17 routines in the planned catalog (16 scheduled plus 1 triggered pipeline).

Session scope:

• Batch 2: 5 Claude Code routines (content_gap_run, ai_visibility_audit, competitor_scrape, internal_link_opportunity, new_content_pipeline)
• Batch 3: 3 GitHub routines (schema_validator, backlink_check, redirect_chain_check), 2 local Mac routines (outlook_triage, screaming_frog_audit), and 10 SKILL.md stubs

The trigger that made this session work where the earlier all-in-one attempt failed: explicit batching. Build 4-6 things, verify, package, deliver, pause. The discipline that got Batch 1 delivered by Claude Code got Batches 2 and 3 delivered by Cowork.

Context Decisions Established Early

Before writing Batch 2, Sam confirmed the last open question: for new_content_pipeline, the routine runs fully autonomous through research, outline, draft, design, SEO, and QA stages, then writes to WordPress as status=draft. Never publish. Sam flips to publish manually in wp-admin after review.

This answer shaped the pipeline's safety architecture, not just a configuration flag.

What Was Accomplished

Batch 2: Claude Code Routines (9 files, 27KB zip delivered)

content_gap_run (Wednesday 06:00 ET)

• Loads firm-briefing, content-gap-hunter, brand-voice-enforcer
• Reads state/competitor-content.json (produced Tuesday by competitor_scrape)
• Cross-references SEMrush keyword gap + GSC current rankings + prior Thursday's AI citation data
• 5-factor scoring: keyword volume, competitive gap, AI citation gap, brand fit, local relevance
• Outputs top 10 scored topic candidates to Notion with 2-sentence briefs, statuses pending
• Never auto-creates WordPress drafts (research only)

ai_visibility_audit (Thursday 06:00 ET)

• Loads firm-briefing, ai-visibility-auditor