The Agent Skills Directory

[SAFE]: The skill is entirely composed of Markdown documentation and evaluation JSON files. It does not contain any executable scripts (Python, JavaScript, shell), configuration files for package managers, or binary assets.
[SAFE]: All external URL references target authoritative and well-known documentation sources for web accessibility, specifically w3.org (W3C) and a11ysupport.io. These are documented neutrally as intended resources for developer reference.
[PROMPT_INJECTION]: No malicious prompt injection patterns, role-play overrides, or safety filter bypasses were found. The instructional language in the metadata and reference files is appropriate for a specialized knowledge base and focuses on triggering the agent for accessibility-related queries.
[DATA_EXFILTRATION]: No evidence of hardcoded credentials, sensitive file path access, or unauthorized network operations. The skill lacks the capabilities to perform file system or network actions.
[OBFUSCATION]: Scanned all content for Base64 encoding, zero-width characters, homoglyphs, and hidden text patterns. All content is plain text and clearly readable, with no signs of obfuscation.
[REMOTE_CODE_EXECUTION]: No remote code execution patterns, package installations, or dynamic code generation were identified.
[METADATA_POISONING]: Metadata fields (name, description, author) accurately reflect the content and purpose of the skill. No deceptive instructions are embedded in the metadata.
[INDIRECT_PROMPT_INJECTION]: While the skill processes user queries about accessibility, it lacks exploitable capability sinks (such as subprocess calls or file writing). The risk surface for indirect prompt injection is negligible, as it only returns informational content from its own Markdown references.

aria-patterns