Campaign Brief Builder Agent A new AI agent called Campaign Brief Builder Agent turns user inputs into structured marketing campaign briefs, flagging missing information and assumptions for human review. The agent is governed by an open-source AgentAz specification that enforces read-only access, cost limits, and human handoffs, ensuring it never fabricates data or guarantees results. Overview Turns your inputs into a structured campaign brief: objective, audience, channels, messaging, KPIs, timeline. Grounds everything in what you provide and marks missing pieces as TBD. Flags assumptions and decisions that need a marketing lead. Defensive: never fabricates audience stats or benchmarks, and never guarantees results. AgentAz™ specification A lightweight, design-time governance spec for security review. It documents what this agent is authorized to do — and why — and pairs with whatever policy engine you already run. It does not enforce anything at runtime. Machine-readable contract agentaz.json , validated against the open AgentAz™ JSON Schema — bundled for offline use and published at a permanent URL: { "$schema": "./agentaz.schema.json", "version": "2.0.0", "last reviewed": "2026-06-24", "agent id": "campaign-brief-agent", "trust level": "A2", "dna pattern": "Synthesis", "worst case action": "Produces a brief with a flawed assumption for lead review. No fabricated stats; results never guaranteed.", "authority boundary": "Builds briefs grounded in inputs; marks gaps TBD; launch/budget tools absent; no guarantees.", "tags": "marketing", "campaign-brief", "read-only", "human-review" , "tool boundary": { "allowed tools": "read inputs", "structure brief", "mark tbd", "flag assumptions" , "execution tools absent": true }, "output boundary": { "format": "structured json", "never emits": "launch", "commit budget" , "never fabricates": true, "estimates not guarantees": true }, "cost boundary": { "max usd per trace loop": 0.2, "alert threshold usd": 0.14 }, "loop boundary": { "max reasoning turns": 8 }, "human handoff": { "triggers": "missing input", "assumption made" , "destination": "marketing lead" }, "audit": { "append only": true, "logs": "inputs", "assumptions" } } New to this? Read the AgentAz specification guide /agentaz-specifications — Trust Levels, DNA patterns, and how it complements your runtime. AgentAz™ is open source under Apache-2.0 https://www.apache.org/licenses/LICENSE-2.0 — schema frozen v1.0.0 and source on GitHub https://github.com/agent-kits/agentaz . Governance matrix A scannable summary of this blueprint's governance coverage, derived from its AgentAz™ specification. It documents the boundaries that already ship — not new functionality. | Agent goal | Bounded by the authority spec above | |---|---| | Trust Level | A2 — Recommend | | Tool access | Least privilege — execution tools absent read-only | | Context handling | Grounded in provided inputs; cites or flags rather than guessing | | Memory strategy | Task-scoped; no persistent cross-session memory | | Human approval | Required on missing input, assumption made → marketing lead | | Audit trail | Append-only log inputs, assumptions | | Cost & loop bounds | ≤ $0.2 per loop · ≤ 8 reasoning turns | | Recovery / escalation | Escalates to marketing lead | Agent component mapping A framework-neutral view of how this blueprint maps to standard agent-architecture components the vocabulary common to ADK-style frameworks . It describes structure for clarity — not an official integration or certified compatibility. | Agent | Primary reasoner — Recommend authority A2 | |---|---| | Tools | read inputs, structure brief, mark tbd, flag assumptions — execution tools absent read-only | | Memory | Task-scoped working context; no persistent cross-session memory | | Guardrails | Worst-case classified A2 ; no execution tools; ≤ $0.2/loop · ≤ 8 turns | | Evaluator | Confidence and authority-boundary checks; low-confidence or out-of-bounds results are flagged, not actioned | | Handoff | Escalates to marketing lead on missing input, assumption made | Failure modes Specific ways this blueprint can fail, and how it is designed to detect, contain, and recover from each — the boundaries that make it safe to run, stated plainly. Includes a fabricated audience statistic or benchmark. - Detection - Figures are tied to provided inputs and unsupported numbers are flagged. - Mitigation - It never fabricates stats; missing data is marked TBD and there are no launch or budget tools. - Recovery - The lead supplies real data or removes the claim. Implies a guaranteed result, such as a conversion lift. - Detection - Guarantee language is flagged. - Mitigation - It states estimates, never guarantees. - Recovery - The lead reframes or removes it. Builds on a flawed or missing assumption silently. - Detection - Assumptions and missing inputs are surfaced as TBD. - Mitigation - The brief flags decisions for the lead. - Recovery - The lead fills the gaps before approval. Evaluation Factual grounding and the absence of fabricated stats or guarantees are primary — an invented benchmark or an implied guarantee is the failure. | Grounding rate | Share of figures and claims traceable to provided inputs. | |---|---| | Fabrication rate | Frequency of invented stats or benchmarks — should be near zero. | | Guarantee-language rate | Frequency of guarantee phrasing rather than estimates — should be near zero. | | Completeness | Share of required brief sections present, with gaps marked TBD. | | Acceptance rate | Share of briefs a lead approves with little change. | Recommended approach. Use briefs with known inputs and reference outputs; measure grounding and audit for fabricated stats and guarantee language. Missing inputs must appear as TBD, not invented. When to use Use it when - You want a structured first-draft campaign brief fast from rough inputs. - You want assumptions and missing inputs surfaced, not hidden. - You want KPIs and audience data grounded in real input or marked TBD. - You want a starting brief a marketing lead refines, not auto-committed spend. Avoid it when - You want it to invent audience data or benchmarks to look complete — it won't. - You expect it to guarantee results or commit budget. - You have no campaign inputs for it to ground the brief in. - You need a final approved brief without human review. System prompt You are a Campaign Brief Builder Agent. You turn provided inputs into a structured marketing campaign brief for a team to refine. You are judged on a useful, well-structured brief and on never fabricating data, guaranteeing results, or committing budget. == CORE PRINCIPLES == 1. Ground in the inputs. Build the brief from what's provided objective, audience, channels, budget, timeline . Don't invent audience segments, market data, or numbers that weren't given. 2. Honest about gaps. Where information is missing — KPIs, budget, audience detail — mark it TBD and flag it, rather than filling it with invented figures. A good brief shows what's undecided. 3. Recommend, don't promise or commit. Suggest channels, messaging, and KPIs. Never guarantee results/ROI, and never commit spend — those are decisions for the marketing lead/stakeholders. == HARD RULES NON-NEGOTIABLE == - NO FABRICATED DATA: Never invent audience statistics, market size, benchmark conversion/CTR numbers, or competitor data. Use provided values or mark 'TBD — needs data'. - NO RESULT GUARANTEES: Never promise a specific ROI, reach, or conversion outcome. Targets are goals/estimates, clearly caveated. - NO BUDGET COMMITMENT: Propose budget allocation as a recommendation; don't commit or finalize spend. - FLAG ASSUMPTIONS & DECISIONS: Label assumptions and surface choices needing a marketing lead/stakeholder. - INFERRED vs PROVIDED: Distinguish what you inferred from what was actually provided. == METHOD == - Take the inputs. Draft the brief sections grounded in them. Mark missing KPIs/audience/budget as TBD. Flag assumptions and decisions. Caveat any projections. == OUTPUT FORMAT return ONE JSON object == { "objective": "