45ck root-cause-hypothesis-generator
Purpose
Generate plausible root-cause hypotheses from symptoms, recent changes, and system context without overstating certainty.
Trigger this skill when
- A maintenance, support, defect, release, or evolution decision needs clearer structure.
- Symptoms, causes, impacts, and release choices are being conflated.
- You need artifacts that reduce operational surprise and rework.
Expected inputs
- issue report or change request
- system context
- known symptoms or requested change
- constraints / urgency
- available evidence if any
Deliverables
- summary
- findings
- structured outputs
- risks / blockers
- open questions
- recommended next skill
Operating procedure
- Clarify whether the work is driven by a defect, environmental change, requested improvement, or preventive maintenance need.
- Separate observations, assumptions, hypotheses, and verified facts.
- Produce the skill-specific artifact or decision aid.
- Surface blast radius, rollback needs, regression exposure, and consumer impact explicitly.
- Recommend the next best handoff instead of trying to solve the entire maintenance lifecycle at once.
Quality gates
- The output is specific to the current maintenance context.
- Uncertainty, evidence gaps, and operational constraints are visible.
- The result is usable by an engineer, maintainer, on-call responder, or team lead.
- The output improves safety, speed, or clarity of the next maintenance decision.
Handoff targets
- change-impact-mapper
- regression-risk-reviewer
- rollback-plan-writer
Output style
- Prefer operationally useful artifacts over generic “maintenance best practices.”
- Be explicit about urgency, blast radius, rollback risk, and residual uncertainty.
- Surface what still needs verification instead of implying the cause is proven.
- Keep outputs concise enough to drop into tickets, incident notes, PRs, or runbooks.
Failure modes to avoid
- Do not confuse a symptom with a verified root cause.
- Do not recommend shipping changes without considering rollback and regression exposure.
- Do not hide environment or data dependencies that could invalidate the result.
- Do not treat all maintenance work as “just bug fixing.”
Minimum output skeleton
## Summary
## Findings
## Structured outputs
## Risks / blockers
## Open questions
## Recommended next skill