What task are you trying to do?
Help users create their own PawWork instructions when they want personalized behavior, without making PawWork silently generate a default AGENTS.md on their behalf. Users should understand that AGENTS.md is their editable preference and project-context file, not an invisible factory default.
What do you do today?
Issue #230 is moving the PawWork factory baseline toward the bundled system prompt and away from default-generated AGENTS.md files. That keeps the runtime contract cleaner, but it leaves a product gap: users who want custom working style, confirmation rules, writing preferences, project conventions, or personal boundaries need an obvious path to create their own AGENTS.md. Today that path is discoverable only if the user already knows the AGENTS.md convention.
What would a good result look like?
PawWork offers a lightweight onboarding or Settings entry point for creating user-owned instructions. The flow should explain the difference between PawWork default behavior, project instructions, and user customization; offer a concise English starter template; and let the user choose whether to create project-local AGENTS.md or a PawWork global instruction file. It should not be required for first-run success, and it should not block issue #230.
Which audience does this matter to most?
Both
Extra context
Follow-up to #230. #230 should handle the runtime baseline by strengthening the PawWork system prompt and removing global Claude fallback. This issue owns the separate UX layer that helps users intentionally create their own AGENTS.md later. Keep this out of #230 acceptance criteria so the runtime contract and onboarding UI can ship independently.
What task are you trying to do?
Help users create their own PawWork instructions when they want personalized behavior, without making PawWork silently generate a default AGENTS.md on their behalf. Users should understand that AGENTS.md is their editable preference and project-context file, not an invisible factory default.
What do you do today?
Issue #230 is moving the PawWork factory baseline toward the bundled system prompt and away from default-generated AGENTS.md files. That keeps the runtime contract cleaner, but it leaves a product gap: users who want custom working style, confirmation rules, writing preferences, project conventions, or personal boundaries need an obvious path to create their own AGENTS.md. Today that path is discoverable only if the user already knows the AGENTS.md convention.
What would a good result look like?
PawWork offers a lightweight onboarding or Settings entry point for creating user-owned instructions. The flow should explain the difference between PawWork default behavior, project instructions, and user customization; offer a concise English starter template; and let the user choose whether to create project-local AGENTS.md or a PawWork global instruction file. It should not be required for first-run success, and it should not block issue #230.
Which audience does this matter to most?
Both
Extra context
Follow-up to #230. #230 should handle the runtime baseline by strengthening the PawWork system prompt and removing global Claude fallback. This issue owns the separate UX layer that helps users intentionally create their own AGENTS.md later. Keep this out of #230 acceptance criteria so the runtime contract and onboarding UI can ship independently.