fsm-protocol-guard

name	fsm-protocol-guard
description	Reference skill for high-risk Room FSM and WebSocket protocol review support. Used by strategy_orchestrator, contract_design_reviewer, contract_auditor, or implementation_auditor to identify high-risk FSM/protocol work, impacted invariants, and required validation. This skill does not replace stage/routing/audit decisions and does not by itself authorize implementation or completion.

FSM Protocol Guard

Overview

Use this skill only as a reference aid for strategy_orchestrator, contract_design_reviewer, contract_auditor, or implementation_auditor when a task may touch high-risk room flow behavior. Return risk evidence and invariant impact only; do not expand into execution ownership or implementation instructions.

This skill does not:

• replace stage classification
• replace team-shape selection
• replace contract review or implementation audit
• authorize implementation by itself
• declare the task complete by itself

It only helps identify:

• whether the task belongs to a high-risk FSM/protocol surface
• which invariants are affected
• which validation scope is required
• what residual risk must remain visible

Inputs

• User request or bounded change summary
• Current task framing from the orchestrating/reviewing agent
• Expected affected files and layers
• Existing contract sources:
  • AGENTS.md
  • WORKFLOW.md
  • QUALITY.md
  • docs/design/01_fsm.md
  • docs/design/02_ws_protocol.md
  • docs/design/03_data_model.md
  • docs/design/07_constants.md

Invoke this skill only after the calling agent has framed the task enough to judge whether FSM/protocol risk may exist.

Workflow

1. Detect whether the change touches FSM/protocol risk areas with references/fsm-protocol-risk-matrix.md.
2. Determine whether repository rules imply:
   • high-risk treatment
   • Plan Mode recommendation
3. Check invariant impact with references/fsm-protocol-invariants.md.
4. If contract behavior changes are implied, identify which design docs must be updated first.
5. Produce only:
   • risk recommendation
   • impacted invariants
   • required validation groups
   • residual risks or unresolved uncertainty

Do not:

• create downstream handoff packets
• proceed into implementation by yourself
• declare the task complete by yourself
• replace orchestrator/auditor judgment with skill-local verdicts

Decision Rules

• Recommend high-risk treatment when the task touches FSM transitions, timer/deadline semantics, authoritative acceptance rules, idempotency behavior, public lobby summary behavior, failure/recovery behavior, or WS schema/payload contracts.
• Recommend Plan Mode when one or more plan-required triggers match.
• If uncertain, recommend the safer high-risk/Plan Mode path.

These rules support orchestration and review. They do not replace orchestrator or auditor ownership of final decisions.

Output Format

Risk recommendation: <high-risk / not high-risk> Plan Mode recommendation: <required / not required> Matched triggers:

• <trigger 1>
• <trigger 2>

Invariants impacted:

• : preserved | tightened | changed
  • Note:

Required validation groups:

Residual risks:

References

• references/fsm-protocol-risk-matrix.md
• references/fsm-protocol-invariants.md
• references/fsm-protocol-checklist-template.md