메뉴
Validates recent code changes with ruthless, direct criticism. Uses git diff, Kubernetes/infrastructure docs (cc), and real-world search (exa) to find bugs, misconfigurations, missing edge cases, and anti-patterns. Use after any commit or before merging.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
Browser automation via Browser Use Cloud API. Handles full lifecycle: account creation via challenge-response signup, session dispatch, polling, and result extraction. Use when user needs web browsing, scraping, flight search, price comparison, form filling, or any task requiring a real browser agent. Trigger: "browser use", "browse", "use browser agent", "automate browser", or any task needing real browser interaction beyond simple page fetch.
Explore mode with built-in critic. Combines openspec-explore thinking with on-demand design validation via sub-agent. Use when the user wants to explore ideas AND get critical feedback without switching modes. Trigger with /explore-critic or "explore with critic".
Ingest a Claude Code or pi-agent session into the llm-wiki knowledge base. Parses JSONL session files, extracts clean conversation transcript, drops it into raw/sessions/ for standard wiki ingest. Use when user says "/memory", "save session", "ingest session", or "remember this session".
Analyze stocks to find discounted/undervalued opportunities using fundamental metrics
| name | validate-change |
| description | Validates recent code changes with ruthless, direct criticism. Uses git diff, Kubernetes/infrastructure docs (cc), and real-world search (exa) to find bugs, misconfigurations, missing edge cases, and anti-patterns. Use after any commit or before merging. |
Ruthlessly validate the most recent change. No hand-holding. No sugar-coating. Find what's wrong.
# Default: last commit
git show --stat HEAD
git diff HEAD~1..HEAD
# Or if user specifies a range
git diff <base>..<head>
Read the diff completely. Understand every line changed.
Determine what the change touches:
Use cc_resolve-library-id then cc_query-docs to check:
Example for Kubernetes changes:
cc_resolve-library-id: query="Kubernetes CronJob concurrencyPolicy", libraryName="kubernetes"
cc_query-docs: libraryId=<resolved-id>, query="CronJob spec concurrencyPolicy behavior and edge cases"
Use exa_web_search_exa to find:
Query examples:
"Kubernetes CronJob concurrencyPolicy Replace gotchas production issues""Helm template race condition known issues"If highlights are insufficient, follow up with exa_crawling_exa on the most relevant URLs.
Structure the output as:
## Verdict: [PASS | FAIL | RISKY]
### Problems Found
[Numbered list. Each item: what's wrong, why it matters, evidence/link]
### Missing Considerations
[What the change didn't account for. Edge cases, failure modes, race conditions]
### Questions for the Author
[Things that can't be determined from code alone — intent, environment, constraints]
Rules for the verdict:
This skill only reports findings. It does not fix, edit, or change any files.
concurrencyPolicy: Allow)concurrencyPolicy: Allow)helm template . --debug)startingDeadlineSeconds set for CronJobsactiveDeadlineSeconds set for Jobs that shouldn't run foreverset -e or proper error handling"$var" not $var)activeDeadlineSeconds on Jobs (can run forever)startingDeadlineSeconds on CronJobs (missed schedules pile up)