Stand up the campaign-setup agent in code, wired to read a brief and emit a draft campaign — but with platform calls stubbed. This proves the loop end to end before we depend on real platform APIs.
In scope:
- Agent process and entry point
- Brief input → draft campaign output
- Stubbed integration layer that records the call it would have made
- Memory write of action log entry
- Audit log entry with rationale
Out of scope (separate issues):
- Real platform integration
- Eval harness
- HITL UX
Depends on: #1, #2, #3, #6, #7, #8, #9 reaching at least skeleton agreement.
Stand up the campaign-setup agent in code, wired to read a brief and emit a draft campaign — but with platform calls stubbed. This proves the loop end to end before we depend on real platform APIs.
In scope:
Out of scope (separate issues):
Depends on: #1, #2, #3, #6, #7, #8, #9 reaching at least skeleton agreement.