What problem does it solve? Writing changesets in a monorepo is error-prone: contributors pick the wrong package, choose the wrong semver bump, or write changelog entries for changes users cannot perceive. This Skill encodes the pythinker-code repository's changeset conventions so every entry is consistent and user-facing. ## Core Features & Use Cases - Write-or-skip decision: Determines whether a change deserves a changeset at all, skipping docs-only, test-only, and internal refactors that users cannot perceive. - Package and bump selection: Maps changes to the correct package (@pymodel/pythinker-code for user-facing changes, internal packages otherwise) and picks patch/minor, with major bumps requiring explicit approval. - Wording and workflow rules: Enforces one short user-facing sentence per changeset, kebab-case filenames under .changeset/, and a confirmation step before committing. - Use Case: After fixing a bug where tool call results were lost in long conversations, run this Skill to produce .changeset/tidy-pandas-fix.md listing @pymodel/pythinker-code as a patch with a single clear sentence. ## Quick Start Ask the agent to generate a changeset for the current uncommitted changes in the pythinker-code repository.
