The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill reads project source files (HTML, JSX, TSX, Vue) to identify accessibility issues. This creates an attack surface for indirect prompt injection, where an attacker could place malicious instructions in comments or attributes within the code to manipulate the agent's output or reasoning.
Ingestion points: Reads files with extensions .html, .jsx, .tsx, and .vue as specified in SKILL.md.
Boundary markers: None identified to delimit code content from agent instructions.
Capability inventory: Executes grep and npx commands and generates user reports.
Sanitization: None identified for the content being processed.
[COMMAND_EXECUTION]: The skill utilizes shell commands to perform its auditing tasks.
Evidence: Employs grep for static analysis of source files and npx @axe-core/cli for dynamic testing.
Context: These commands are limited to local static analysis and testing against localhost.
[EXTERNAL_DOWNLOADS]: The skill uses npx to fetch and execute external tools.
Evidence: Runs npx @axe-core/cli@4 http://localhost:3000 to perform audits.
Context: The download targets a well-known and widely used accessibility testing package from the npm registry.

a11y-audit