Content system

This page explains how CRISP is designed to produce reliable outcomes: clear intent, a human approval gate, and deterministic publishing behavior.

Who this page is for

• Operators who need to understand “why it works this way”
• Support/admins troubleshooting edge cases
• Reviewers validating that publishing is safe and deterministic

Intent-driven prompting (philosophy)

CRISP content generation is built around intent and constraints:

• We specify the “why” (goal, audience, offer)
• We preserve brand voice rules and guardrails
• We generate channel-specific drafts (LinkedIn vs 𝕏 vs Meta vs Blog)

Approval workflow

• Drafts arrive in the approval queue.
• A human can edit and approve.
• Approval is the gate that allows a platform publish flow to start (where supported).

Scheduling logic and queue guards

CRISP uses queue guards to reduce platform risk and user surprise:

• Minimum spacing rules can be enforced per destination (example: Meta Phase 1 uses per-destination spacing when creating jobs).
• Some platforms are export-only and do not support scheduling yet (𝕏 threads, Blog).

Status model (publish jobs)

Native publishing uses job states that make failures explicit:

• queued → publishing → published
• queued/retrying → failed (after max attempts)

Retries and failure states

• Retries use backoff to avoid hammering platform APIs.
• Permanent failures are surfaced clearly so a human can take action.

Related docs

• Publishing overview: Publishing overview
• Security & privacy: Security & privacy