claude-plugin

Structured bug investigation using formal models for hypothesis-driven debugging, with hashharness MCP storage as the append-only backend (experimental variant of formal-debugger). Builds scoped Alloy models (normative rules, data constraints, causal chains, observability) to generate distinguishing experiments and narrow the causality cone around a symptom. Use whenever the user reports a bug, incident, or unexpected behavior and wants rigorous investigation backed by hashharness's append-only hash-chained storage. Trigger on "investigate", "root cause", "debug this", "why does this happen", "find the bug", "postmortem", or any unexplained symptom. Valuable for: business-rule bugs, state machine errors, workflow inconsistencies, cross-service disagreements, investigations where the obvious explanation was already checked. Not for simple stack traces or typos — use when the problem space is large enough that unstructured search would waste time or miss the real cause.

EXPERIMENTAL — UNDER TESTING. Unstable variant; behaviour and gates may change without backwards-compatible migration. Structured bug investigation using textual formal models for hypothesis-driven debugging, with hashharness MCP storage as the append-only backend (textual-only variant of formal-debugger, no solver execution). Builds scoped textual models (state machines, sequence diagrams, schemas, invariant tables — see `references/representation-menu.md`) covering normative rules, data constraints, causal chains, and observability, then walks them by hand against collected evidence. Use whenever the user reports a bug, incident, or unexpected behavior and wants rigorous investigation backed by hashharness's append-only hash-chained storage AND wants to avoid Alloy/Dafny tooling. Trigger on "investigate", "root cause", "debug this", "why does this happen", "find the bug", "postmortem", or any unexplained symptom. Valuable for: business-rule bugs, state machine errors, workflow inconsistencies, cross-service

Structured bug investigation using formal models for hypothesis-driven debugging. Builds scoped Alloy models (normative rules, data constraints, causal chains, observability) to generate distinguishing experiments and narrow the causality cone around a symptom. Use whenever the user reports a bug, incident, or unexpected behavior and wants rigorous investigation. Trigger on "investigate", "root cause", "debug this", "why does this happen", "find the bug", "postmortem", or any unexplained symptom. Valuable for: business-rule bugs, state machine errors, workflow inconsistencies, cross-service disagreements, investigations where the obvious explanation was already checked. Not for simple stack traces or typos — use when the problem space is large enough that unstructured search would waste time or miss the real cause.

Write, run, and interpret Alloy 6 formal models for software systems, business processes, and skill/workflow design. Use this skill whenever the user asks to model, verify, or prove properties about state machines, lifecycle flows, data invariants, API contracts, integration boundaries, permission systems, or business rules. Also trigger when the user mentions Alloy, formal methods, formal verification, model checking, counterexample, temporal logic, or wants to "prove" something about their system. Use even for seemingly simple state-machine questions — the solver catches edge cases that humans miss. This skill applies to any domain: SaaS, e-commerce, fintech, DevOps, or business process design.