Apply shared planning baseline (format, verifiability, assumptions) before creating any concrete change plan
Install
npx skillscat add ozerohax/assistagents/planning-base Install via the SkillsCat registry.
SKILL.md
Provide a universal change plan skeleton: short, verifiable, no hidden assumptions
Define quality rules: what a good plan looks like and what not to do
Always before producing a change plan
When the task is ambiguous and questions/assumptions must be captured
</when_to_use>
Separate facts, assumptions, and decisions
The plan must be verifiable: every step can be confirmed by a test/observation
Changes are incremental, with stop points
Do not replace requirements with solutions: first "what", then "how"
Write only relevant sections: do not fill the template for the sake of it
</core_principles>
Goal
Constraints
Proposed change
Risks
Verification
Rollout / rollback (if needed)
Open questions
</default_output_format>
The goal is stated as behavior/outcome, not as a list of actions
Acceptance criteria exist (what "done" means)
Boundaries are stated: what we explicitly do NOT do in this task
A verification plan exists (at least: one quick signal + one regression check)
Open questions are explicitly listed and affect the plan
</quality_rules>
Do not produce a plan built on hidden assumptions
Do not mix independent goals into one change-set without a reason
Do not add "nice-to-have improvements" without a clear goal
</do_not>
Goal: remove 500s on /orders. Proposed change: add a guard for null tax and a regression test. Verification: the repro is gone, the test is green. Risks: may affect tax calculation. Open questions: do we have legacy clients without tax?
Has a goal, a concrete change, verification, risks, and questions
Do something and check manually.
No verifiability and no specifics