The Agent Skills Directory

[SAFE]: The skill's design and its use of potentially sensitive tools like Bash and file system editors are commensurate with its stated purpose of auditing and repairing code for accessibility compliance.
[COMMAND_EXECUTION]: The skill is configured to use the Bash tool, which is used for standard repository navigation and tool orchestration as part of the auditing and fixing workflow.
[EXTERNAL_DOWNLOADS]: The skill performs live accessibility audits by interacting with external URLs. These operations are conducted through specialized browser-based tools and are essential for rendering-level auditing.
[PROMPT_INJECTION]: The skill processes untrusted data from both local files and external web pages, creating a surface for indirect prompt injection. The skill mitigates this by instructing the agent to treat the output of specialized auditing tools as authoritative when performing mechanical fixes.
Ingestion points: External DOM content accessed via audit_live and audit_browser_collect; local source code files accessed via Read, Glob, and Grep.
Boundary markers: The instructions do not define specific delimiters for untrusted input processing.
Capability inventory: The skill can modify the file system via Write and Edit, execute shell commands via Bash, and coordinate tasks through Skill and Task.
Sanitization: No specific content sanitization is described, but the agent's reliance on structured tool directives for code modifications limits the scope of influence from external content.

audit