FG-0513 Anti-Spin Protocol
Description
When an agent is blocked, the correct response is to escalate collaboration, not escalate force. The complexity gradient has a speed limit, and that speed limit has a human checkpoint. One failed attempt at the obvious approach, then stop and check in. Pitch the complex solution before executing it. This prevents the pattern where agents burn hours on increasingly elaborate workarounds for problems that a thirty-second human conversation would resolve.
Implementation
Define a clear escalation sequence: (1) try the obvious approach, (2) if it fails, describe the problem and proposed alternatives to the human, (3) the human selects or redirects, (4) execute the approved approach. Never attempt more than one complex workaround without human approval. Track incidents where force was escalated instead of collaboration, and review them to refine the protocol.
Success Indicators
- Blocked tasks are resolved through conversation rather than brute force
- Time-to-resolution for blocked tasks decreases over time
- Agents present options rather than executing increasingly desperate workarounds
- The human is consulted at decision points, not informed of decisions already made
Failure Modes
- Agent pride: treating escalation as failure rather than protocol, leading to increasingly creative but unproductive attempts
- Over-escalation: asking the human about trivially solvable problems, degrading trust and wasting human attention
- Spin disguised as progress: the agent reports 'trying alternative approaches' while burning cycles on variations of the same failed strategy
- Human unavailability: the escalation target is not responsive, and the agent has no fallback protocol
Sub-Methods 3
One failed attempt at the obvious approach, then check in with the human operator. Do not attempt a second complex approach without approval. The first failure contains diagnostic information that the human needs to see. Presenting 'I tried X and it failed because Y, here are options Z1 and Z2' is always more productive than 'I tried X, Y, Z1, Z2, and Z3, they all failed, what now?'
Complex solutions must be proposed to the human before execution begins. Describe the approach, the expected time investment, the risk, and the alternatives. The human may know a simpler path. The human may know the approach will not work. The human may redirect entirely. Any of these outcomes saves more time than executing a complex solution that turns out to be wrong.
Recognize the signs of spinning: multiple attempts at variations of the same approach, increasing complexity without increasing results, time investment growing while confidence shrinking. When you detect spin in yourself or your agent, invoke the protocol immediately. Spin is not persistence. It is the sunk cost fallacy applied to problem-solving.