The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user or agent to download and install external code from the repository github.com/Community-Access/accessibility-agents using the npx skills add command. This repository is not verified.
[COMMAND_EXECUTION]: The skill provides explicit shell commands (npx skills add) for the agent to execute in order to modify its runtime environment with external capabilities.
[DYNAMIC_CONTEXT_INJECTION]: The skill documentation includes the ! command syntax (e.g., ! npx skills add), which is interpreted by some agent platforms as a directive to execute shell commands immediately upon the skill being loaded or referenced in chat.
[INDIRECT_PROMPT_INJECTION]: This skill is specifically designed to audit and process potentially untrusted external data sources including HTML, ARIA structures, PDFs, Word documents, and Excel spreadsheets. 1. Ingestion points: Web code specialist agents and document accessibility specialists (Word/Excel/PDF) ingest external file content. 2. Boundary markers: The stub contains no instructions for the agent to use delimiters or to ignore instructions embedded within the audited files. 3. Capability inventory: The skill performs audits and generates remediation guidance, requiring it to read and interpret complex file structures. 4. Sanitization: No sanitization or validation logic is specified for the ingested document content.

accessibility-agents