The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted source code files (HTML, React, Vue, Svelte, Astro) provided by the user to perform accessibility audits. This creates a vulnerability where malicious instructions embedded in these files (e.g., in HTML comments or string literals) could attempt to override the agent's primary instructions.
Ingestion points: The skill uses the Read tool to ingest the content of target frontend files as specified in the SKILL.md logic.
Boundary markers: The instructions do not explicitly mandate the use of delimiters or 'ignore' instructions to separate the untrusted file content from the skill's operational logic.
Capability inventory: The skill is granted Edit and Write permissions, which allows the agent to modify the filesystem based on the results of the (potentially manipulated) analysis.
Sanitization: No sanitization or filtering of the ingested file content is performed before the agent evaluates it against the accessibility rules.

front-a11y