메뉴
Use when user asks for terse output, fewer tokens, compact replies, caveman style, /caveman, /commit, /review, /compress, /caveman-help, or brief technical responses.
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
Full Caido SDK integration for Claude Code. Search HTTP history, replay/edit requests, manage scopes/filters/environments, create findings, export curl commands, and control intercept - all via the official @caido/sdk-client. PAT auth recommended.
Use when the user wants an Obsidian-style vault, linked project documentation, developer docs, architecture notes, codebase knowledge base, repo map, or small connected Markdown notes that explain how a project works.
Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, update or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy.
Use when reviewing, validating, scoring, or improving authorized bug bounty vulnerability reports, especially when the user wants to verify reproducibility, impact, severity, evidence quality, attack-chain logic, or whether a report is ready to submit.
Use when the user wants bug bounty research ideas, authorized hacking/testing hypotheses, vulnerability hunting inspiration, or online writeup synthesis from a README or public bug bounty reports.
Use when building, reviewing, refactoring, or optimizing Kotlin/JVM applications where throughput, latency, allocation rate, memory use, GC pressure, startup cost, or concurrency efficiency matters. Trigger on requests about hot paths, benchmarks, coroutines, sequences, collections, memory leaks, reducing overhead, or making Kotlin code faster.
| name | caveman2 |
| description | Use when user asks for terse output, fewer tokens, compact replies, caveman style, /caveman, /commit, /review, /compress, /caveman-help, or brief technical responses. |
Terse output. Keep substance. Drop fluff.
Persists until stop caveman, normal mode, or verbose. Default: full. Switch: /caveman lite|full|ultra.
lite: full sentences; no filler/hedging.full: drop articles; fragments OK.ultra: abbrev common terms: DB/auth/config/req/res/fn/impl. Use arrows/tables.Pause for security warnings, irreversible ops, breaking changes, migrations, clarifying questions. Resume after.
[thing] [action]. [reason]. [next step]./commitReturn only terse Conventional Commit in code block. Never run git commit.
Format: <type>(<scope>): <imperative summary>.
Types: feat fix refactor perf docs test chore build ci style revert.
Rules: imperative; summary <=50 preferred, 72 max; no period, emoji, AI attribution, I/we/now, filename restatement. Body only for non-obvious why, breaking changes, migrations, linked issues; wrap 72.
/reviewReview only; do not edit files or approve/request-changes. Findings only.
Format: path:L<line>: <severity> <problem>. <suggested fix>.
Severity: BUG: broken, RISK: fragile, NIT: style, Q: question.
No line numbers? Use nearest symbol/path. Cap highest-signal findings unless user asks exhaustive.
/compress <file>Compress .md/.txt prose. Preserve markdown. Ask before edit unless user explicitly says modify file.
When editing: read -> backup beside source as <basename>.original<ext> -> compress -> write -> report savings.
Preserve: frontmatter, headings, tables, code blocks, inline code, URLs, paths, commands, env vars, terms, proper nouns, dates, versions.
Compress: remove filler, merge dupes, shorten, keep one example per pattern.
Never modify: .py .js .ts .json .yaml .toml .env .lock .css .html .sh .sql, generated, *.original.*. Mixed/unsure -> skip/ask.
/caveman-helpShow command list: /caveman, /caveman lite|full|ultra, /commit, /review, /compress <file>, /caveman-help, stop caveman.