Go Project

When an AI agent resumes work on a project that has an existing docs/PLAN.md and docs/TODO.md, the agent must read these files first, understand the current state, execute the next task, and update the documentation to reflect progress. This skill ensures continuity between sessions and prevents the agent from re-deriving the plan or deviating from established decisions.

Workflow

When the user asks you to continue, implement, or make progress on a project, follow this workflow:

1. Read existing docs — Read docs/PLAN.md, docs/TODO.md, and any files in docs/references/ if they exist. Understand the design decisions already made and what tasks remain.
2. Confirm understanding — Briefly summarize to yourself: what is the project, what was the last completed step, and what is the next immediate step?
3. Execute the next task — Implement the task described in TODO.md as the "Next immediate step." Follow the design in PLAN.md. Do not deviate from established decisions unless the user explicitly requests it.
4. Update TODO.md — Mark completed tasks with [x]. Reveal the next task as the new "Next immediate step." If new tasks were discovered during implementation, add them. Remove the "Next immediate step" section and recreate it at the top with the new next task.
5. Update PLAN.md if needed — If design decisions changed during implementation (e.g., a library didn't work as expected and an alternative was chosen), update PLAN.md to reflect the new decision. Do NOT add new decisions that the user didn't confirm.
6. Save research to references/ — If you looked up an API, library, or technical detail during implementation, save the findings to docs/references/ instead of keeping them in context. Follow the same guidelines as the create-project skill: write in markdown, include the source at the top, summarize key findings, and note trade-offs or caveats.
7. Summarize in conversation — Report what was done, what was completed, and what the next step is.

Important Rules