| name | prompt-template-definition |
| description | Define and manage structured multi-line prompt templates for AI agents, specifically for SWE-bench tasks and multimedia skill extraction.
|
| license | MIT |
| metadata | {"version":"1.0"} |
| triggers | ["prompt template","SWE_BENCH_PROMPT_TEMPLATE","MMSkillTrainer_PROMPT_TEMPLATE"] |
Prompt Template Definition
This skill involves defining high-level instructions for AI agents using structured Python strings.
SWE-bench Problem Solving Template
When creating a prompt for solving software engineering tasks, include:
- A placeholder for the
{problem_description}.
- Specific instructions for outputting the solution (e.g.,
/workspace/solution.md).
- Logic for skill extraction: identifying reusable components and saving them in a specific
SKILL.md format under /workspace/skills/<skill_name>/.
Multimedia Skill Extraction Template
When creating a prompt for summarizing multimedia into skills:
- Instruct the assistant to analyze files and identify demonstrated skills.
- Mandate the use of tools (
list_skills, read_skill) to check for existing duplicates.
- Enforce a specific XML-based output format (
<skill_name> and <skill_description>) containing YAML frontmatter and Markdown content.
Formatting Requirements
- Use triple quotes
""" for multi-line strings.
- Ensure the YAML frontmatter includes:
name, description, license, and triggers.
- Use snake_case or hyphen-separated names for variables and skill files.
