Forum Moderation & Content Extraction Plan

Overview

Two connected objectives built on top of the Forum Content Ingestion System (System 3, live since 2026-04-07):

1. Moderation — Detect off-topic or guideline-violating posts, flag them for human review, and eventually automate removal + polite email redirection
2. Content extraction — Analyze on-topic posts for insights that inform programming decisions, curriculum development, and participant understanding

Both depend on the forum content already flowing into the vault via the ingestion system.

Status: Planning. Requires Asia’s review before implementation begins.

---

Why

Moderation

Participants occasionally post content to the forums that belongs in a private email exchange — billing questions, receipt requests, concerns about terms and conditions, or reactions to media usage. These posts:

• Are visible to the entire group, which is not the appropriate venue
• Can set a tone where the forum drifts from its educational purpose
• Require manual monitoring and response, which doesn’t scale

Existing examples already in the vault:

• Receipt/invoice request (Any Guay, 2026-03-27) — billing question posted publicly
• Media consent concern (triggered by Clementine’s group feed comment, 2026-03-14) — Patrick already responded with a redirect to email for admin matters

The Forum Guidelines and Media Consent post both explicitly state: administrative, logistical, billing, and consent matters should go to email, not the forum.

Content extraction

On-topic posts contain valuable signal: what participants find confusing, what resonates, which forms or concepts generate the most discussion, what questions come up between sessions. This information should feed back into programming decisions without requiring manual review of every thread.

---

Part 1 — Forum Moderation

What gets flagged

A post is off-topic if it falls outside the scope defined in the Forum Guidelines. Categories:

Category	Examples	Action
Billing/payment	Receipt requests, invoice questions, payment issues	Flag → remove → email
Terms/policy concerns	Questions about terms of service, refund policy, participation waiver	Flag → remove → email
Media/consent concerns	Objections to photo usage, privacy concerns	Flag → remove → email
Promotions/endorsements	Promoting other products, services, personal offerings	Flag → remove → warning email
Off-topic personal	Unrelated personal content, trolling	Flag → human review → case-by-case

Posts by admins (Patrick, Asia) are never flagged — admin posts may address these topics as public responses or redirections.

Workflow — Phase 1 (human-in-the-loop)

Forum content sync pulls new topics/replies (every 2 hours)
    ↓
Classification step in forum-content-sync.py
    ↓ (checks each new post against guideline categories)
Flagged post → notification sent (email or Slack to Patrick/Asia)
    ↓
Human reviews the flag + suggested response
    ↓
Human approves → system sends email to poster + removes post from forum
    ↓
Removed post archived in vault with status: moderated

Classification approach:

• Keyword/pattern matching first (billing, payment, receipt, invoice, facture, recu, refund, terms, waiver, consent, photo, privacy, promotion, advertis-)
• If keyword match: flag with category and confidence
• Future: Claude-based classification for ambiguous cases (Phase 2 dependency — not designed now)

Notification: Email to Patrick and Asia with:

• Post title, author, group, date
• Matched category and keywords
• Link to the post on practice.baseworks.com
• Suggested email response (templated per category)
• One-click approve/reject (email reply or simple web form — TBD)

Email templates (one per category)

Each template follows these principles:

• Acknowledge the person’s concern as legitimate
• Explain that the forum is for course-related discussion only
• Confirm the post has been removed from the forum
• State that the conversation continues via email
• Provide a direct way to reply

Example — billing/payment:

Thank you for reaching out about your [receipt/invoice/payment].

The group forum is reserved for discussions related to the course content and practice. For billing, payment, and account questions, we handle those directly by email so we can address them properly and privately.

Your forum post has been removed, and this email thread is now the place to continue. [Specific response to their question follows.]

Templates to be drafted for all five categories. Asia to review tone and French translations.

Post removal mechanism

• BuddyBoss REST API: POST /buddyboss/v1/topics/{id} with status: trash (or spam)
• Same SSH → wp eval pattern as the ingestion system (Cloudflare workaround)
• Vault file updated: frontmatter gets status: moderated, moderation-category, moderation-date, moderation-action fields
• Post body preserved in vault for record-keeping — only removed from the live forum

Workflow — Phase 2 (automated with human override)

Once templates are validated and the pattern is proven:

• Classification runs automatically on each sync
• If confidence is high (exact keyword match + known category): auto-send email + auto-remove post
• If confidence is low (ambiguous): flag for human review as in Phase 1
• All actions logged in vault with full audit trail
• Weekly digest email summarizing all moderation actions

Phase 2 is not designed beyond this outline. It depends on Phase 1 proving the categories and templates work.

---

Part 2 — Content Extraction

What gets extracted

From on-topic posts (those that pass moderation), extract structured insights:

Signal type	What to look for	How it helps
Confusion points	Questions about forms, concepts, or instructions	Identifies where curriculum explanations need refinement
Resonance	Posts describing breakthroughs, “aha” moments, or positive shifts	Validates what’s working — useful for communications and programming
Form-specific discussion	Posts tagged with or mentioning specific forms	Maps which forms generate the most engagement or difficulty
Between-session practice	Reports of self-guided practice, questions about home practice	Indicates engagement depth and where support is needed
Concept connections	Participants linking Baseworks concepts to other domains	Shows how the method is being understood and contextualized

Output format

A periodic extraction report (suggested: after each sync cycle or weekly digest) written to:

02-areas/practice-platform/community-forums-groups/extraction/
    index.md
    {YYYY-MM}-extraction-report.md    ← monthly

Each report contains:

• Summary of new posts in the period
• Categorized insights (confusion points, resonance, form-specific, etc.)
• Links back to the source topic files in the vault
• Suggested action items (e.g., “3 participants asked about Star Tilt mechanics — consider addressing in next session”)

Implementation approach

• Phase 1: Manual — Claude reads new topic files during session prep and surfaces relevant insights (the /session-prep skill, when built, would incorporate this)
• Phase 2: Automated tagging in forum-content-sync.py — each topic gets extraction-tags in frontmatter based on content analysis
• Phase 3: Monthly extraction reports generated automatically

Phase 1 is available immediately — the data is already in the vault. Phases 2-3 are not designed beyond this outline.

---

Implementation Steps — Moderation Phase 1

1. Draft email templates for all five categories (EN + FR) — Asia review required
2. Add classification function to forum-content-sync.py — keyword/pattern matching
3. Add notification mechanism (email via WP wp_mail or direct SMTP from VPS)
4. Add post-removal function (BuddyBoss API status: trash via SSH → wp eval)
5. Add moderation-* frontmatter fields to vault file update logic
6. Create approval mechanism (email reply parsing or simple web endpoint)
7. Test with historical examples (receipt post, media consent trigger)
8. Deploy with human-in-the-loop — all flags require manual approval

---

Existing Forum Guidelines (reference)

The classification system enforces rules already published to participants:

From Forum Guidelines:

• Use relevant tags aligned with learning content
• Share insights related to Primer Course content only
• Do not endorse or promote other products or personal offerings
• Posts with out-of-context endorsements or promotions will be removed
• Violations may result in a ban from the platform and course

From Media Consent post:

• “The group feed and forum are spaces for sharing observations, questions, and discussions about your experience with the practice and the program content”
• “For administrative or logistical questions — including participation terms, consent, billing, or similar matters — we’d ask that you reach out to us by email”

From Forum Tag Shortlist:

• All tags are course-content related (forms, concepts, session references, methodology)
• No administrative or billing tags exist — reinforces that such topics don’t belong

---

Dependencies

• Forum Content Ingestion System (System 3) — live, provides the data
• Email sending capability from VPS (WP wp_mail via SSH or direct SMTP)
• Asia’s review of email templates and moderation categories
• Human approval workflow (mechanism TBD — email reply or web form)

---

Open questions for Asia

1. Are the five moderation categories complete, or are there other types of off-topic posts you’ve seen?
2. Should the email response come from a specific address (e.g., info@baseworks.com, pat@baseworks.com)?
3. For the content extraction: what specific signals would be most useful for programming decisions?
4. Should moderated posts be completely removed from the vault, or kept with a moderated status for the record?
5. Any adjustments to the email template tone/approach?

---

Related

• Forum Content Ingestion Plan — the ingestion system this builds on
• Vault Sync Systems — architecture pattern
• Forum Guidelines — the rules being enforced
• Forum Tag Shortlist — approved content tags
• Community Forums & Groups — vault index
• Practice Site Platform Infrastructure