What problem does it solve?
It turns algorithmically complex internal interfaces described in tech.yaml into language-agnostic reference pseudocode, so teams can implement deterministic behavior instead of guessing at tricky flows.
Core Features & Use Cases
- Detects algorithmic complexity: Scans internal_interfaces behavior_contracts for complexity indicator patterns and also supports an explicit algorithm_spec_required: true trigger.
- Produces reference pseudocode per interface: Generates reference-algorithms.md with one section per qualifying interface, including complexity classification, pseudocode, invariants, and edge cases.
- Gates output behind human approval: Writes a DRAFT artifact and requires a user tether action to lock it before evidence is recorded.
Quick Start
Run /algorithm for your issue branch that contains a prepared {stm_base}/{issue}/context/design/tech.yaml describing the complex interfaces you want pseudocode for.