promptfoo-strategies-indirect-injectionlisted

# Indirect Prompt Injection Strategy (`indirect-web-pwn`) Tests whether AI agents with web browsing capabilities can be manipulated through malicious instructions embedded in web pages. The strategy dynamically generates realistic web pages containing hidden attack payloads, adapted to your target's purpose. Plugin-agnostic — works with any plugin to test data exfiltration and behavior manipulation. ## Use this skill when - Testing AI agents with web browsing/fetch capabilities - Testing if injected instructions in web pages can trick the agent into leaking data - Testing if injected instructions can manipulate agent behavior or output - Testing data exfiltration to external URLs (with `data-exfil` plugin) - Testing indirect prompt injection via fetched web content (with any other plugin) - Layering jailbreak strategies with web-based injection for more effective attacks - Testing multi-turn attacks with embedding rotation ## Do not use this skill when - Testing RAG context injection (retrieved docs) → use `promptfoo-redteam-rag` (`indirect-prompt-injection` plugin) - Testing static encoding bypasses → use `promptfoo-strategies-static` - Testing dynamic iterative refinement → use `promptfoo-strategies-dynamic` - Testing multi-turn conversation attacks → use `promptfoo-strategies-multi-turn` - Testing MCP tool poisoning → use `promptfoo-redteam-agents` - The target agent cannot fetch URLs (no web browsing capability) ## 🛡️ Edge cases (mandatory handling) - **Target must h