adversarial-prompting
abe238 作
AI & LLMs概要
Adversarial analysis to critique, fix.
説明
このスキルでできること
Applies a 7-phase adversarial methodology to complex problems. Generates multiple solution approaches, critiques each for weaknesses, develops and validates fixes, then consolidates into ranked recommendations. Works for technical problems like architecture decisions or debugging, and non-technical problems like business strategy.
Forces systematic identification of failure modes and edge cases before implementation, catching problems that single-pass analysis typically misses.
使いどころ
- Choosing between competing architecture patterns before committing
- Identifying failure modes in a proposed security design
- Stress-testing a product strategy for overlooked risks
- Debugging a complex performance issue with multiple possible causes
- Evaluating tradeoffs in a high-stakes infrastructure migration
ワークフロー例
User asks: Should I migrate our monolith to microservices?
- 1
Generates 3-5 distinct migration approaches (full rewrite, strangler fig, domain decomposition, etc.)
- 2
Critiques each approach for failure modes: operational complexity, data consistency risks, team capacity gaps
- 3
Develops specific fixes for each identified weakness
- 4
Validates each fix to confirm it resolves the root cause without introducing new problems
- 5
Consolidates into ranked options with trade-off summaries and a final recommendation
Ranked list of viable migration strategies with a concrete recommendation, implementation steps, and early warning indicators to monitor
インストール
clawhub install abe238/adversarial-promptingCLIコマンドを実行するか、Claude Code、Cursor、Codexなどお使いのコーディングエージェントにプロンプトを貼り付けてください。