Regulated & High-Stakes Workflows

Safe, evidence-backed
civic advocacy workflows.

Coordinate specialized agents to turn a community concern into a verified campaign packet — with provenance, safety review, and human-action tasks before anything goes public.

Band coordinates the agents.

FI preserves memory and provenance.

Safety gates every public action.

Agent flow

Evidenceverified

veto loop

Campaign
Safetyreview
Outreach
Coordinator
Reporterpacket

The high-stakes problem

Advocacy is easy to start and easy to get wrong.

Grassroots groups carry real legal and safety exposure with no ops layer to catch it.

01

Unsupported claims create legal risk.

A defamation suit is the fastest way for a corporation to silence a local group. One overstated accusation is all it takes.

02

Public campaigns can accidentally escalate.

Targeting a person instead of a practice, or leaking private data, turns advocacy into harassment — even when the cause is right.

03

Volunteer teams burn out on coordination.

Groups dissolve because organizers drown in email, spreadsheets, and drafting — not because the cause died.

How the workflow works

Six specialized agents, one governed handoff chain.

Each agent owns one role and exchanges structured context through Band. The coordination is visible — not narrated.

Evidence

verified

Finds and verifies claims, attaching a provenance tier to every source.

Campaign

Turns verified evidence into a campaign narrative — never beyond what the evidence supports.

Safety

veto power

Blocks doxxing, harassment, unsupported accusations, and unsafe escalation before anything ships.

Outreach

Drafts posts, emails, flyers, and public copy — in the language of the community.

Coordinator

Converts the approved strategy into a concrete volunteer task board.

Reporter

packet

Assembles the final provenance and audit packet, including every safety verdict.

The safety veto loop

Safety can veto the campaign before public release.

This is not a prompt chain. It is a governed workflow — and the rejection is on the record.

Campaign draft

“Restaurant X is lying to its customers.”

sent to Safety via Band
Safety verdictvetoed

VETO — unsupported accusation. Requires stronger evidence or softer language.

returned to Campaign for revision
Rewritten & approvedapproved

“Available evidence does not support the restaurant’s ‘compostable’ claim in local disposal conditions.”

The rejected revision ships inside the packet — governance that only shows its approvals is marketing.

The output

A campaign packet, ready for humans to execute.

Hours of specialist work, governed and auditable. The system assembles it; humans hold the send button.

  • Evidence brief
  • Risk-reviewed campaign message
  • Outreach copy
  • Volunteer task plan
  • Source / provenance report
  • Safety audit log

The architecture

Why Band + FI + Safety.

Band

Coordinates agent handoffs, shared context, and workflow state — the visible coordination layer.

FI

Stores provenance, memory, evidence trails, and audit artifacts across the whole run.

Safety

Reviews every public-facing action before release. Legal-risk-aware by design.

Insult AI cross-examined claims. Activist OS coordinates evidence-backed action.

Turn a community concern into safe, coordinated action.

First use case: vegan and ecological community campaigns, including greenwashing detection and local action planning.