The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructions direct the agent to install multiple sub-skills from a community GitHub repository (mastepanoski/claude-skills) using the npx skills add command. This creates a dependency on unverified third-party code that is not from a trusted organization or well-known service.\n- [COMMAND_EXECUTION]: The skill requires the agent to execute shell commands in the local environment to detect the installation scope and perform the installation of the sub-skills (e.g., npx skills add ..., [ -f "$f" ] && ...).\n- [METADATA_POISONING]: There is a branding discrepancy where the skill is attributed to one author identity (podo) while the content and resources belong to another identity (mastepanoski), which can be deceptive regarding the origin and trustworthiness of the code.\n- [INDIRECT_PROMPT_INJECTION]: The skill is designed to perform UX, UI, and accessibility audits, which inherently involves the agent processing external, potentially attacker-controlled project data.\n
Ingestion points: Processes external UI designs, accessibility code, and information architecture documents (e.g., sitemaps).\n
Boundary markers: No specific delimiters or safety instructions are provided to the agent to mitigate risks from malicious instructions embedded in the data being audited.\n
Capability inventory: The agent is granted shell execution capabilities for installation and file system access for environment verification.\n
Sanitization: The instructions lack any sanitization or validation steps for the external content before it is processed by the agent.

mastepanoski-skills