The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill functions by ingesting untrusted target URLs provided by the user to perform security assessments, which exposes the agent to potential instructions embedded in remote content.
Ingestion points: Target URLs and IP addresses provided by the user for scanning (SKILL.md).
Boundary markers: Explicit 'Mandatory confirmation gate' instructions require the agent to verify scope and authorization with the user before performing any probing or exploitation.
Capability inventory: Performs network operations via curl to inspect HTTP response headers and provides templates for local HTML proof-of-concept testing.
Sanitization: The skill relies on human-in-the-loop confirmation rather than automated input sanitization for the target URLs.

hunt-clickjacking