The Agent Conflict Preventer

Design conflict prevention for parallel agents.

Agents running in parallel: [DESCRIBE]
Shared resources: [FILES / DATABASE / API / STATE]
What they each modify: [DESCRIBE]

Produce:
1. Every place two agents could touch the same thing
2. For each: what goes wrong if they do
3. Prevention strategies: partitioning, locking, ordering,
   append-only design, or isolation
4. Which strategy fits each conflict, and why
5. Detection: how would I notice a conflict occurred
6. Resolution: what happens when one is detected
7. What to tell each agent about the others' existence

The best answer is usually partitioning — give each agent a
disjoint slice of the work so conflicts are structurally
impossible. Tell me whether that is achievable here, and if not,
why not.

Flag any design where two agents write to the same file. That is
the single most common multi-agent failure and it usually
manifests as one agent silently undoing another's work.