"## Available Skills\n\n<usage>\nSkills provide specialized capabilities and domain knowledge. Use them when they match your current task.\n\nHow to use skills:\n- Load skill: load_skill_through_path(skillId=\"<skill-id>\", path=\"SKILL.md\")\n- The skill will be activated and its documentation loaded with detailed instructions\n- Additional resources (scripts, assets, references) can be loaded using the same tool with different paths\n\nExample:\n1. User asks to analyze data \u2192 find a matching skill below (e.g. <skill-id>data-analysis_builtin</skill-id>)\n2. Load it: load_skill_through_path(skillId=\"data-analysis_builtin\", path=\"SKILL.md\")\n3. Follow the instructions returned by the skill\n\nMetadata is rendered as XML under each <skill> element:\n- scalar metadata becomes a simple child element\n- nested maps become nested XML elements\n- lists become repeated <item> elements\n- <skill-id> is always appended for tool loading\n</usage>\n\n<available_skills>\n\n"
"\n## Code Execution\n\n<code_execution>\nYou have access to the execute_shell_command tool. When a task can be accomplished by running \na pre-deployed skill script, you MUST execute it yourself using execute_shell_command rather \nthan describing or suggesting commands to the user.\n\nSkills root directory: %s\nEach skill\'s files are located under a subdirectory named by its <skill-id>:\n %s/<skill-id>/scripts/\n %s/<skill-id>/assets/\n\nWorkflow:\n1. After loading a skill, use ls to explore its directory structure and discover available scripts/assets\n2. Once you find the right script, execute it immediately with its absolute path\n3. If execution fails, diagnose and retry \u2014 do not fall back to describing the command\n\nRules:\n- Always use absolute paths when executing scripts\n- If a script exists for the task, run it directly \u2014 do not rewrite its logic inline\n- If asset/data files exist for the task, read them directly \u2014 do not recreate them\n\nExample:\n # Explore what scripts are available for a skill\n execute_shell_command(command=\"ls %s/data-analysis_builtin/scripts/\")\n\n # Run an existing script with absolute path\n execute_shell_command(command=\"python3 %s/data-analysis_builtin/scripts/analyze.py\")\n</code_execution>\n"
"\n## Code Execution\n\n<code_execution>\nYou have access to the execute_shell_command tool. Each skill in <available_skills>\nincludes a <files-root> element giving the absolute path to that skill\'s files.\n\nWorkflow:\n1. After loading a skill, look at its <files-root> in <available_skills>\n2. List its files: ls <files-root>/\n3. Run scripts: python3 <files-root>/scripts/<script-name>\n4. Always use absolute paths derived from <files-root>; never invent paths\n5. If a script exists for the task, run it directly \u2014 do not rewrite its logic inline\n</code_execution>\n"