From kivi-claude-skills
Causal-chain debugging that traces from symptoms to root cause BEFORE making any code changes. Prevents shallow fixes and wrong initial approaches.
How this skill is triggered — by the user, by Claude, or both
Slash command
/kivi-claude-skills:deepfixThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Causal-chain debugging that traces from symptoms to root cause BEFORE making any code changes. Prevents shallow fixes and wrong initial approaches.
Causal-chain debugging that traces from symptoms to root cause BEFORE making any code changes. Prevents shallow fixes and wrong initial approaches.
DO NOT EDIT ANY FILE UNTIL STEP 3 IS COMPLETE.
This is not a suggestion — it is a hard constraint. Steps 1 and 2 are READ-ONLY.
Build the full chain before proposing any fix:
[Symptom] ← [Proximate Cause] ← [Deeper Cause] ← [Root Cause]
Example:
"ReviewList shows 0 items"
← API returns empty array
← Query uses status="IN_REVIEW" but data has "in_review"
← Frontend/backend enum convention mismatch (no shared contract)
Mandatory cross-module checks:
Tech-stack specific checks:
| Stack | What to check |
|---|---|
| Go + GORM | gorm:"column:..." tags, json:"..." tags, binding:"..." tags, AutoMigrate vs actual DB schema |
| Python + FastAPI | Pydantic alias, response_model, async/await correctness, SQLAlchemy column names |
| Vue 3 + TS | Prop types vs API response shape, .value unwrapping, Element Plus event names (@change vs @update:modelValue) |
| Docker | Env var names consistency across .env, docker-compose.yml, and app config |
Present to the user:
Wait for user acknowledgment if confidence is MEDIUM or LOW.
/deepfix — start debugging the current issue (user describes the bug)/deepfix <error message> — start from a specific error/deepfix <file:line> — start from a specific locationtry/catch to suppress an error without understanding why it occursnpx claudepluginhub phoxiao/kivi-claude-skills --plugin kivi-claude-skillsApplies systematic root-cause analysis to debug errors, test failures, and unexpected behavior across frontend, backend, database, network, and performance.
Traces root causes of errors, stack traces, test failures, and build issues by gathering evidence, following call chains backward, and explaining the full causation narrative in structured format.
Guides systematic root cause analysis for bugs, test failures, unexpected behavior, and technical issues before implementing fixes.