Use when an agent should remember to ask Shellbrain for targeted recall at task start, subproblem changes, repeated failures, hypothesis changes, or closeout.
Use when an agent should remember to ask Shellbrain for targeted recall at task start, subproblem changes, repeated failures, hypothesis changes, or closeout.
Define and enforce the owner's language-agnostic clean-core architecture for new software projects and refactors. Use when deciding responsibility placement, dependency direction, structural stage, safe refactor moves, semantic types, pure policies, entrypoints, startup wiring, adapters, and dependency boundaries.
Clean existing code by enforcing semantic integrity, reader clarity, and local efficiency without changing architecture. Use when auditing or improving code that has fake optionality, silent fallbacks, stale compatibility, permissive parsing, unclear names, dead wrappers, redundant work, or over-abstraction. Prefer required typed state, explicit errors, fail-fast validation, deletion of obsolete branches, clear names, simple functions, and focused tests.
Use when a reviewer agent needs one fast cross-repo picture of how agents are using Shellbrain, where the product is working, where it is failing, and which expected capabilities are being skipped.
Use when a reviewer agent needs one fast cross-repo picture of how agents are using Shellbrain, where the product is working, where it is failing, and which expected capabilities are being skipped.