Skip to content

Latest commit

 

History

History
29 lines (23 loc) · 3.48 KB

File metadata and controls

29 lines (23 loc) · 3.48 KB

Principles

The decision-forcing layer of the .codebuddy. Principles are leaf skills: small, one-idea files that agents must apply by name when a decision point matches them. When a workflow or review changes a decision, the agent must name the principle that changed it.

Principles are thin and always loaded conceptually — they are not workflows, they are the why behind decisions. Workflows and agents reference them by name (principle: prove-it-works).

Principle Core Idea
never-block-on-the-human Proceed on reversible work; present the result, let the human course-correct.
prove-it-works Verify against the real artifact, not "it compiles" or "looks right".
boundary-discipline Guards at system boundaries; trust internal types; keep business logic pure.
experience-first Choose user delight over implementation convenience.
model-the-domain Encode the domain in a structure instead of scattered conditionals.
make-operations-idempotent Converge to the same end state, no matter how many times the operation runs.
fix-root-causes Trace each symptom to its root cause. Reproduce first, ask why until you reach it.
outcome-oriented-execution Converge on the target architecture; don't preserve throwaway compatibility states.
laziness-protocol Bias to deletion and the smallest change that solves the problem.
guard-the-context-window Route bulk to subagents; keep summaries in the main thread.
migrate-callers-then-delete-legacy-apis Introduce the new API, migrate all callers, delete the legacy one in one wave.
sequence-verifiable-units Break work into small units that each end in a check. Verify each before the next.
subtract-before-you-add Remove dead weight first, then build on the simpler base.

How They Enforce

  • Agents declare their strongest principles in ### 1a. Wiring and must apply them by name.
  • Code review's PATTERN REVIEW pass checks that a change still honors the project's principles.
  • A decision that goes against a principle must be flagged explicitly, not silently bent.

Related: Agents · Skills · Root README