The Agent Skills Directory

[PERSISTENCE]: The skill is designed to create and modify persistent instruction files in vscode-userdata:/User/prompts/ and .github/instructions/. These locations are used by the agent to define its behavior across different sessions and projects. While this is the primary purpose of the skill, it establishes a mechanism to permanently alter the agent's operating instructions.
[INDIRECT_PROMPT_INJECTION]: The skill ingests 'lesson content' from user input or current chat sessions and writes it into persistent memory files without sanitization.
Ingestion points: The skill parses user-provided lesson content and analyzes chat session history to extract learnings.
Boundary markers: There are no specified delimiters or 'ignore embedded instructions' warnings used when writing the extracted lessons into the resulting .instructions.md files.
Capability inventory: The skill performs file-write operations to global (vscode-userdata:/User/prompts/) and workspace-specific (.github/instructions/) instruction directories.
Sanitization: The process does not include validation or sanitization of the content being 'remembered'. If the agent is currently processing malicious data (e.g., analyzing a file with hidden prompt injection) and is triggered to /remember a lesson from it, the malicious instructions could be saved to the user's global or project-level instruction set.

remember