Skip to content

README.md

Latest commit

 

History

History
319 lines (258 loc) · 9.92 KB

README.md

File metadata and controls

319 lines (258 loc) · 9.92 KB
Raw

SimpliXio Automation (Monorepo)

This folder is the automation engine for proof, weekly learning, and warm acquisition loops. It includes acquisition research + drafting automation with compliance gates.

It is monorepo-aware:

  • reads source data from the main repo (../growth_output, ../weekly_digest_*.md)
  • writes automation artifacts to cortexos_automation_scripts/output/

What It Does (Simple Words)

Proof automation

The proof scripts do this:

  1. Collect and filter real project signals.
  2. Build real product artifacts (SimpliXio Today, Weekly Review, Decision Replay).
  3. Generate draft posts (X, LinkedIn, blog), Discord proof drafts, HTML pages, and cards from those artifacts.
  4. Run quality checks (no hype words, no weak/generic copy, no repeated angle).
  5. Only queue publish output if quality passes and publish flags are enabled.

Important:

  • By default this is safe/dry-run.
  • It does not invent traction/users/revenue.
  • It writes logs and summaries every run.

Acquisition / prospection automation

The acquisition scripts do this:

  1. Collect public, compliant lead signals (GitHub, RSS, HN, internal artifacts).
  2. Score lead fit for SimpliXio (fit, candidate, not_fit).
  3. Draft personalized private outreach messages for high-fit leads.
  4. Run acquisition quality/compliance checks.
  5. Save everything in SQLite CRM + JSON/Markdown reports.

Important:

  • Private outreach is draft-only by default (needs_approval).
  • No LinkedIn scraping.
  • No automatic cold DM/email sending by default.
  • Public content queueing still requires quality pass + explicit publish flags.

What Is Automatic vs Manual

Fully automated:

  • lead discovery
  • lead scoring
  • outreach draft generation
  • public content draft generation
  • quality gate checks
  • run logs + summaries

Manual approval required:

  • private outbound sending (DM/email/comment/reply)
  • any high-risk public posting choice

Trust Defaults

  • Private by default.
  • Public drafts must pass quality gates before queueing.
  • No LinkedIn scraping or automated LinkedIn activity.
  • No automatic cold outreach sends in this repo.
  • Human approval stays required for private outbound.

Structure

cortexos_automation_scripts/
  marketing_automation.py
  scripts/
    filter_signals.py
    build_cortex_today.py
    build_weekly_review.py
    build_decision_replay.py
    build_public_newsletter.py
    build_discord_proof_drafts.py
    generate_newsletter.py
    build_decision_examples.py
    build_public_proof_archive.py
    marketing_quality_gate.py
    publish_outputs.py
    run_weekly_pipeline.py
    acquisition_crm.py
    lead_collector.py
    lead_scorer.py
    outreach_drafter.py
    content_engine.py
    acquisition_quality_gate.py
    run_acquisition_pipeline.py
  output/
    cortex_today/
      cortex_today.json
      cortex_today.md
      cortex_today.html
      archive/
    filtered_signals/
    weekly_review/
      latest.json
      latest.md
      latest.html
      archive/
    decision_replay/
      latest.json
      latest.md
      latest.html
      archive/
    newsletters/
      drafts/
      approved/
      rejected/
      logs/
      latest.json
    discord/
      release_notes_latest.md
      build_in_public_latest.md
      weekly_review_latest.md
      product_lesson_latest.md
      feedback_prompt_latest.md
      latest.json
    decision_examples/
      # runtime exports may be added later
    drafts/
    quality_gate/
    logs/

Local setup

cd cortexos_automation_scripts
python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
cp .env.example .env

Daily run order (artifact-first)

python3 scripts/filter_signals.py
python3 scripts/build_cortex_today.py
python3 scripts/build_weekly_review.py
python3 scripts/build_decision_replay.py
python3 scripts/generate_newsletter.py --period weekly --mode weekly-lessons --strict-safety --strict-taste
python3 marketing_automation.py
python3 scripts/marketing_quality_gate.py --strict
python3 scripts/build_discord_proof_drafts.py
python3 scripts/publish_outputs.py

Single command (recommended):

python3 scripts/run_weekly_pipeline.py --strict-quality

Weekly pipeline

python3 scripts/run_weekly_pipeline.py --strict-quality

Decision Examples

Decision Examples are the public-safe long-tail acquisition layer. They are committed under ../docs/decision-examples/ so they can be reviewed, linked, and published without depending on runtime output.

Validate:

python3 scripts/build_decision_examples.py --check

Rebuild the committed Markdown archive:

python3 scripts/build_decision_examples.py

Rules:

  • synthetic examples must be clearly marked as examples
  • every example must include exactly 3 priorities
  • every priority needs a concrete why and action
  • sensitive/private material is rejected
  • no fake traction, fake users, or fake metrics
  • no thin SEO pages

Public Proof Archive

The public proof archive packages safe proof surfaces into one Desire Loop:

capture messy input -> 3 priorities -> why -> action -> public-safe proof

Rebuild:

python3 scripts/build_public_proof_archive.py

Committed outputs:

  • ../docs/public-proof/index.md
  • ../docs/public-proof/proof_manifest.json
  • ../docs/public-proof/changelog/README.md
  • ../docs/public-proof/weekly-review/README.md
  • ../docs/public-proof/decision-replay/README.md
  • ../docs/public-proof/newsletter-examples/README.md

Fastest way (recommended)

From repo root (/Users/pierre/Code/CortexOSLLM), use the Makefile wrappers:

make autopilot-weekly
make autopilot-acq-daily
make autopilot-acq-weekly
make autopilot-all

Why this helps:

  • avoids forgetting command order
  • keeps strict quality checks on by default
  • reduces copy/paste mistakes during releases

Canonical runbook:

  • /Users/pierre/Code/CortexOSLLM/cortexos_automation_scripts/AUTOMATION_RUNBOOK.md

The pipeline writes:

  • JSON run log: output/logs/weekly-pipeline-*.json
  • Markdown summary: output/summaries/weekly-pipeline-*.md
  • Curated Discord drafts: output/discord/latest.json + output/discord/*_latest.md
  • Discord channel coverage: release-notes, build-in-public, weekly-review, product-lessons, feedback

Acquisition automation

Daily:

python3 scripts/run_acquisition_pipeline.py --mode daily --strict-quality

Weekly:

python3 scripts/run_acquisition_pipeline.py --mode weekly

Acquisition outputs:

  • SQLite CRM: output/acquisition/acquisition.sqlite3
  • Raw lead signals: output/acquisition/raw/lead_signals_*.json
  • Lead shortlist: output/acquisition/drafts/latest_lead_shortlist.md
  • Outreach drafts: output/acquisition/drafts/latest_outreach.md
  • Acquisition quality report: output/acquisition/quality_report.json
  • Pipeline logs: output/acquisition/logs/acquisition-*.json
  • Pipeline summaries: output/acquisition/summaries/acquisition-*.md

Lead scoring tiers:

  • fit: high-confidence prospect, drafted for manual approval
  • candidate: near-threshold prospect, held for manual review (not drafted automatically)
  • not_fit: archived for now

Safety defaults:

  • private outreach is always saved as needs_approval
  • no LinkedIn scraping
  • no outbound sending in these scripts
  • public publish queue requires PUBLISH_PUBLIC=true and quality pass
  • Discord output is draft-only and requires manual posting.

Single command outputs (daily):

  • JSON log: output/acquisition/logs/acquisition-daily-*.json
  • Markdown summary: output/acquisition/summaries/acquisition-daily-*.md
  • Lead shortlist: output/acquisition/drafts/latest_lead_shortlist.md
  • Outreach drafts: output/acquisition/drafts/latest_outreach.md

Newsletter generation

Weekly public-safe newsletter draft:

python3 scripts/generate_newsletter.py --period weekly --mode weekly-lessons --strict-safety --strict-taste

Custom range from selected source IDs:

python3 scripts/generate_newsletter.py \
  --period custom --from 2026-04-01 --to 2026-04-21 \
  --mode product-builder-notes \
  --source-ids thought-12,decision-7 \
  --strict-safety --strict-taste

What it does:

  • reads selected source material (thoughts, notes, decisions, priority-feedback, weekly-review, decision-replay)
  • supports daily, weekly, monthly, or custom date ranges
  • supports source filtering by --source-ids, --tags, and --keywords
  • classifies each source item (private, sensitive, internal, public_safe, public_ready) before generation
  • redacts sensitive patterns (emails, phone numbers, long numbers, URLs, common secret token formats)
  • blocks confidential indicators and marks unsafe drafts as needs_review in strict mode
  • runs a strict taste gate (generic/repetitive/weak drafts become needs_review)
  • updates local voice memory from approved writing only (output/newsletters/voice_profile.json)
  • writes content-flywheel snippets (X, LinkedIn, blog outline, acquisition angle) inside the JSON log
  • stores safety + redaction logs with source IDs for manual review
  • writes Markdown + HTML + JSON metadata with safety + redaction + classification + taste reports
  • archives only safe + taste-passing + public_ready drafts to output/public_proof/newsletters/

Outputs:

  • output/newsletters/drafts/newsletter-*.md (or rejected/ when blocked)
  • output/newsletters/drafts/newsletter-*.html
  • output/newsletters/logs/newsletter-*.json
  • output/newsletters/latest.json
  • output/newsletters/voice_profile.json
  • output/newsletters/memory.json
  • output/public_proof/newsletters/newsletter-*.md (only when archive rules pass)

Notes

  • build_cortex_today.py uses the latest ../growth_output/*/(ready_to_publish|pending_approval).json as primary source.
  • If growth output is missing, it attempts one fallback run of ../scripts/cortex_growth_loop.py.
  • marketing_automation.py generates drafts and artifacts only.
  • publish_outputs.py is safe by default (PUBLISH_DRY_RUN=true), and updates content memory only when quality passes.