Evaluate content against the 4-level approval policy and determine whether to auto-approve, score-gate, send to review, or block. Use when the approval engine needs to make a decision on pending content. Triggers on any content approval evaluation request.
Evaluate content against the AI CMO approval policy and return a structured decision.
qwen2.5-coder:7b — deterministic evaluation, not creative writing.
Read the full policy from: OpenClawData/policies/approval-rules.json
You will receive:
Return EXACTLY this JSON structure:
{
"decision": "<auto_approve|score_gate_pass|score_gate_fail|review|block>",
"level": "<L1|L2|L3|L4>",
"weighted_score": <number>,
"max_dimension": <number>,
"risk_scores": {
"source_confidence": <0-100>,
"brand_voice": <0-100>,
"claim_sensitivity": <0-100>,
"duplication": <0-100>,
"platform_risk": <0-100>,
"data_safety": <0-100>
},
"override_triggered": "<rule name or null>",
"reason": "<1-2 sentence explanation>",
"action": "<approve_and_queue|hold_for_review|block_and_log>",
"destination": "<queues/{channel}/approved/ or approvals/review/ or approvals/blocked/>"
}
Apply in this order (first match wins):
data_safety > 50 → BLOCK. If any override trigger matches → BLOCK.