Recognize and create valuable repository agent skills from concrete repeated workflows. Use automatically after meaningful work when a reusable workflow could prevent future errors or repetition, and when the user asks to add, create, build, or encode a skill.
Recognize and maintain repository agent skills from concrete usage evidence. Use automatically after meaningful work when a skill proves incomplete, incorrect, noisy, unsafe, or stale, and when the user asks to edit, fix, refine, reorganize, or update a skill.
Keep Creed source-code comments essential, durable, and useful to maintainers. Apply whenever adding, rewriting, reviewing, or auditing comments in code, stylesheets, scripts, tests, or migrations.
Create a clean, focused Git commit for completed repository work. Use when the user asks to commit, make a commit, save the work as a commit, or invokes /commit or $commit.
Write, edit, and review Creed user-facing language in its calm, precise, editorial voice. Apply automatically whenever changing product copy, marketing copy, onboarding text, UI labels, buttons, errors, toasts, prompts, documentation, pricing language, or…
Diagnose bugs, regressions, failing checks, unexpected behavior, and performance problems from concrete evidence before changing code. Apply automatically whenever the user asks to debug, investigate, fix a bug, explain a failure, or determine why something…
Keep Creed's standalone user documentation accurate, minimal, and aligned with shipped behavior. Apply automatically when changing apps/docs, setup or doctor flows, deployment and hosting support, configuration or environment variables, agent connections, MCP…
Design, implement, review, and verify safe Supabase Postgres schema migrations for Creed. Apply automatically whenever changing tables, columns, indexes, constraints, functions, triggers, RLS policies, grants, storage policies, or persisted data shapes.