SOP-009 — Atlantis Social Surface Operations

1. Purpose

This Standard Operating Procedure defines how Atlantis ITS operates its social intelligence system across LinkedIn, Bluesky, and Moltbook.

The goal is a controlled social intelligence system that builds reach, captures signal, and routes useful findings into Atlantis operations — without spamming, drifting, or risking brand trust.

This SOP governs day-to-day operations. For AI production risk review before any social system goes live, see SOP-008. For Nova's agent-level behavior rules, see nova_prime_directive_v2.2.md. For the full Phase 3 architecture, see NOVA-PHASE3-SOCIAL-INTELLIGENCE.md.

The Exit Rule (applies to every post on every surface):

If a post does not increase trust, signal, or revenue intelligence — it does not ship.

2. Scope

In scope:

• All Nova social agent actions on Bluesky and Moltbook
• LinkedIn posts drafted by Nova and published manually by Shane or Maranda
• Signal collection, scoring, and routing through Nova Bus and n8n
• Content publishing decisions and approval flows
• Logging and auditability of all social actions
• Cadence management across surfaces

Out of scope:

• Nova Prime Directive ethics enforcement (see nova_prime_directive_v2.2.md)
• AI production risk review before going live (see SOP-008)
• Credential rotation (see SOP-006)
• Platform account creation and initial setup

3. Surface Definitions

Critical distinction: Moltbook is Nova's agent-intelligence surface and public research log. It is not a sandbox runtime. Risky tools, unknown packages, and experimental scripts are tested in the local Docker sandbox on Trunks before any findings are posted to Moltbook. See NOVA-PHASE3-SOCIAL-INTELLIGENCE.md Section 4 for the full sandbox gate.

Moltbook signal trust level: Treat as useful but noisy input. Do not use Moltbook as a single source of truth without corroboration from at least one other source (Scout, Lyra, or verified external reference).

4. Ownership

5. Required Fields

Every item handled by the social system must carry the following fields before routing or publishing:

6. Content Classes

7. Nova Routing Flow

All signals collected from social surfaces follow this routing path:

Signal collected (Bluesky / Moltbook / LinkedIn draft / manual input)
        ↓
Normalize into Required Fields (Section 5)
        ↓
Score: Revenue / Strategy / Novelty
        ↓
Route by score:
  High  → Forgejo issue / Horizon task / Briefing note / Chronicle queue
  Med   → Review queue (Shane or Mercy reviews before action)
  Low   → Archive or drop
        ↓
Publish only if content passes:
  - Content Class check
  - Tone check
  - Disclosure check
  - Shane approval (if required — see Section 9)
        ↓
Log every action (source, score, route, outcome, timestamp)

8. Posting Rules

8.1 LinkedIn

• Posts must be deliberate, professional, and limited to high-value content
• Nova drafts only — Shane or Maranda reviews and publishes manually
• Never auto-post to LinkedIn under any circumstances
• AI involvement must be disclosed where appropriate
• Cadence: 2–4 posts per week maximum

8.2 Bluesky

• Posts may be more frequent but only if structured, concise, and clearly useful
• Every post must carry AI disclosure label
• No raw signal goes public without Nova scoring and class check
• Apply the Exit Rule before every post
• Cadence: 3–8 quality posts daily, only if signal supports it
• Silence is better than noise — never post to fill a cadence slot

8.3 Moltbook

• Read-first, post-second — listening is the primary function
• Do not trust Moltbook as a source of truth without corroboration
• No raw bot chatter goes public without Nova review and sanitization
• Findings from local sandbox testing may be posted only after review
• Cadence: continuous listening, selective capture and posting

8.4 Universal Rules (All Surfaces)

• Disclose AI involvement clearly on all public-facing posts
• Never fabricate data, statistics, or results
• Never astroturf, manufacture consensus, or engage in fake engagement
• Never scrape or store personal data from platform users
• Issue public corrections rather than quiet deletions
• All posts must be traceable to operator (Shane Hardin / Atlantis ITS) and agent

9. Approval Requirements

10. Default Cadence

11. Safety Rules

• Disclose AI involvement clearly on all public-facing posts
• Keep credentials scoped per surface — no shared API keys across surfaces
• Enforce rate limits and publishing caps (see nova_prime_directive_v2.2.md Section 8.4)
• Require human approval for sensitive, reputational, legal, or sales-critical posts
• Pause automation immediately if platform warnings, account flags, or suspicious activity appear — notify Shane via NTFY/Discord alert
• No social agent goes live on any surface without SOP-008 intake completed
• No autonomous posting until Nova Prime Directive is loaded at agent boot

12. Kill Switch

Each social surface must have a documented Kill Switch executable in < 5 minutes.

13. Logging & Auditability

Every social action must be logged. Minimum log fields:

- Timestamp (UTC)
- Surface
- Action type (post / react / comment / DM / signal-capture / route)
- Content class
- Score
- Route destination
- Approval status
- Outcome
- Agent / operator

Log destination: atlantis-ops/intelligence/social-log.json → Forgejo Retention: permanent (never purge without Shane approval)

14. Metrics

Track the following per surface per week:

15. Document Hierarchy

Nova Social is governed by three related documents in strict order of authority. When documents conflict, the stricter safety or control requirement wins.

Prime Directive → SOP-008 → SOP-009 (this doc) → Social Intelligence Architecture

Prime Directive is never overridden. If unsure, apply the stricter rule.

16. Related Documents

• nova_prime_directive_v2.2.md — Nova agent ethics and behavior rules
• NOVA-PHASE3-SOCIAL-INTELLIGENCE.md — Phase 3 architecture, Xavier Field Pattern, signal lanes
• SOP-008 — AI Production Risk Review (run before any social surface goes live)
• SOP-006 — Credential Rotation (rotate App Passwords on schedule)
• atlantis-memory/updates.md — live tactical state
• ATLANTIS_AI_MEMORY.md — long-term canonical memory

17. Revision History

18. Approval

Atlantis ITS · Metro Atlanta · SOP-009 — Atlantis Social Surface Operations Path: atlantis-docs/sops/SOP-009-Atlantis-Social-Surface-Operations-v1.0.md