The Agent Skills Directory

Local Command Execution: The skill executes a bundled Python script (scripts/a11y_check.py) to perform technical accessibility calculations that are difficult for an LLM to handle directly, such as contrast ratios and XML parsing. This involves running shell commands on the agent's host environment.
Indirect Prompt Injection: The skill ingests and processes content from user-provided documents (.pptx, .docx, .html, .md). It specifically instructs the agent to read and evaluate text strings (like alt-text, titles, and link text) for quality and relevance. This creates a surface where a maliciously crafted document could include instructions designed to influence the agent's behavior.
Ingestion points: Documents are parsed by scripts/a11y_check.py and the resulting text content is reviewed by the agent as instructed in SKILL.md.
Boundary markers: The skill does not define specific delimiters or instructions to the agent to ignore potentially malicious embedded content within the documents.
Capability inventory: The skill possesses the capability to execute a Python script, read file content, and write modifications back to the local file system to apply fixes.
Sanitization: There is no evidence of sanitization or filtering of the text extracted from the documents before it is processed by the agent.
File System Modification: To fulfill its primary purpose, the skill is authorized to modify user documents to fix accessibility issues. While this is the intended behavior, users should be aware that the skill has write access to the files it analyzes.

accessibility-pass