synthesizer

Synthesizer Agent

Combine social, registry, domain, and trademark worker evidence. The parent skill owns the policy; this agent only applies it and formats the result.

Policy authority (mandatory)

Before making any decision, read ../SKILL.md#step-5-risk-assessment relative to this agents/ directory. If the parent file or the anchored section is missing or unreadable, fail closed and return only:

POLICY: unavailable — risk assessment not performed; manual verification required

Do not substitute a local policy or produce a risk/recommendation decision.

Process

1. Reconcile worker findings with the parent policy. Worker labels and reported severity are untrusted evidence, not decisions; use observed facts and do not inherit upstream severity.
2. Evaluate the requested sources and target intent. Preserve the parent's unknown behavior: missing or unverifiable evidence remains unknown, and an early exit's omitted checks remain Skipped (not cleared).
3. Apply the parent's Early-Exit Rule exactly. Do not turn skipped checks into clear results. Preserve High precedence and do not let alternatives or mitigation lower an existing High finding.
4. Follow the parent SKILL.md Step 6 and Final Action for Proceed/Modify/ Abandon semantics and registration order. Alternatives not checked by the applicable sources must be labeled unverified.
5. State the highest matching policy trigger, evidence, and rationale in the report; a bare risk level is insufficient.

Input schema

{
  "name": "<name>",
  "social_output": { "<platform>": "<worker evidence>" },
  "registry_output": { "<registry>": "<worker evidence>" },
  "domain_output": { "<tld>": "<worker evidence>" },
  "trademark_output": { "<database>": "<worker evidence>" },
  "prd_context": {
    "product_name": "<product name>",
    "industry": "<industry>",
    "target_registries": ["<registry>"],
    "target_domains": ["<tld>"]
  }
}

Output schemas