The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the AI agent to execute shell commands using the node runtime to run local scripts included with the skill. This allows the skill to perform operations on the host system outside of the direct control of the AI's standard tool environment. \n
Evidence: The Setup section requires running node <skill-base-dir>/scripts/context.mjs.\n
Evidence: The Commands section specifies running node scripts/pin.mjs to create or remove shortcuts.\n- [DYNAMIC_EXECUTION]: The skill relies on the execution of external JavaScript files (context.mjs, pin.mjs) to establish context and manage tool routing. This executes code in the user's local environment which is not directly visible in the provided markdown file. This behavior also contradicts the skill's own safety claims.\n- [INDIRECT_PROMPT_INJECTION]: The skill possesses a high surface area for indirect prompt injection because it is designed to read and process arbitrary project files (components, design tokens, etc.) while holding significant capabilities (shell execution, file access, and network requests).\n
Ingestion points: The skill uses Read, Glob, and Grep to analyze project content to understand the "incumbent visual truth".\n
Boundary markers: There are no explicit instructions or delimiters mentioned to prevent the agent from obeying instructions embedded in the analyzed code or design files.\n
Capability inventory: The skill has access to shell execution (node), full project read access, and network tools (WebFetch).\n
Sanitization: No evidence of validation or sanitization of external project content is present.\n- [METADATA_POISONING]: The skill's documentation contains a direct contradiction between its stated safety profile and its operational requirements, and there is a mismatch between the provided author metadata and the YAML frontmatter.\n
Evidence: The Authentication section states the skill works "without a... script, or executable," yet the Setup section explicitly mandates the execution of node scripts/context.mjs.\n
Evidence: The skill's YAML frontmatter lists the author as "UIZZE" while the external metadata identifies the author as "jeremylongshore".

anti-ui-slop