From gerdsenai
Adversarial analysis of code, documents, repos — security, architecture, dependencies, testing, accessibility with Socratic reasoning chains
How this command is triggered — by the user, by Claude, or both
Slash command
/gerdsenai:red-team <file-or-directory> [--domains code,security,deps] [--depth shallow|standard|deep] [--fix]This command is limited to the following tools:
The summary Claude sees in its command listing — used to decide when to auto-load this command
You are an adversarial analysis engine. Read the full agent protocol at `${CLAUDE_PLUGIN_ROOT}/agents/red-team-reviewer.md` and follow it completely.
The user's target is: `$ARGUMENTS`
Begin at Phase 0 (Tool & Capability Discovery). Do not skip any phase.
## Mode Detection
- If the target is a directory or non-markdown file → full multi-domain analysis
- If the target is a `.md` file → document-scoped analysis (domains: document, strategic)
- If no target provided → ask the user using AskUserQuestion:
- "Analyze a codebase or directory"
- "Review a markdown document"
- "Analyze a ...You are an adversarial analysis engine. Read the full agent protocol at ${CLAUDE_PLUGIN_ROOT}/agents/red-team-reviewer.md and follow it completely.
The user's target is: $ARGUMENTS
Begin at Phase 0 (Tool & Capability Discovery). Do not skip any phase.
.md file → document-scoped analysis (domains: document, strategic)npx claudepluginhub gerdsenai/gerdsenai-markdown-to-pdf-suite-claude-plugin --plugin gerdsenai/red-teamRuns structured red-teaming exercise on an AI feature to identify alignment failures, producing a report with attack surface, misuse scenarios, guardrail gaps, biases, and recommendations.
/red-teamGenerates 10+ adversarial variants of a prompt injection attempt using evasion techniques like homoglyphs and leetspeak, tests safety scanner detection, and reports overall rate plus evaded variants.