The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's core workflow involves the agent ingesting and analyzing content from external, potentially untrusted web pages and the output of automated tools (like Lighthouse and axe-core). This represents a surface for indirect prompt injection where malicious content embedded in a target page could attempt to influence the agent's code-fixing suggestions or overall behavior.
Ingestion points: The skill utilizes lighthouse_audit, take_snapshot, and the axe CLI to fetch and analyze external web content as described in SKILL.md under the 'Evidence-led audit workflow'.
Boundary markers: The instructions do not define specific delimiters or explicit 'ignore embedded instructions' warnings for the content being audited.
Capability inventory: The skill expects the agent to modify the source code ('Fix the source') and perform network operations using CLI tools (axe, lighthouse).
Sanitization: There is no evidence of specific sanitization, filtering, or validation of the data retrieved from external sources before it is processed by the agent.

accessibility