The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes PostToolUse hooks to automatically trigger the execution of local bash scripts (ts-check.sh, auto-prettier.sh, console-log-check.sh) located in the hidden ~/.claude/hooks/ directory. These scripts are executed whenever files with .ts, .tsx, .js, or .jsx extensions are modified, enabling local command execution as a background side effect of agent edits.\n- [PROMPT_INJECTION]: The skill employs behavior-steering language ("CRITICAL PRINCIPLE", "NON-NEGOTIABLE", "MUST NEVER") to override the agent's default logic and mandate the use of documentation-fetching tools. Additionally, it introduces a vulnerability surface for indirect prompt injection by instructing the agent to "Follow fetched docs exactly" for code generation.\n
Ingestion points: Data returned from MCP tools such as fetch_tailwind_docs, fetch_alpine_patterns, fetch_vite_config, and get_ecommerce_ui_pattern.\n
Boundary markers: Absent; the skill does not define delimiters to isolate untrusted external documentation from the agent's instructions.\n
Capability inventory: The skill possesses the ability to modify local files and trigger shell commands via hooks and testing tools (e.g., Playwright).\n
Sanitization: While the skill advises stripping <script> tags from SVGs, it lacks automated sanitization for the textual content ingested from external documentation sources.

frontend-app