The Agent Skills Directory

[SAFE]: The skill is entirely documentation-based, providing best practices, templates, and techniques for prompt engineering.
No executable code, shell scripts, or binaries are included.
Frontmatter allowed-tools (Read, Write) is appropriate for a documentation skill.
[PROMPT_INJECTION]: The skill includes sections on emphasis scales for prompt design (e.g., 'CRITICAL
ZERO TOLERANCE').
These are presented as educational content for prompt engineering and do not attempt to override the current agent's instructions.
[INDIRECT_PROMPT_INJECTION]: The skill provides patterns for handling external data in systems like RAG (Retrieval-Augmented Generation).
Ingestion points: Identified in SKILL.md ([USER_PROVIDED_CONTENT]) and references/rag-patterns.md ({query}).
Boundary markers: The skill explicitly recommends using XML delimiters (e.g., <document>, <context>, <source>) to isolate untrusted data.
Capability inventory: The skill only utilizes Read and Write tools for local file management.
Sanitization: Not applicable, as the skill provides structural patterns for LLM instruction rather than runtime code.

prompt-creation