The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external knowledge base content, including Markdown files and directory structures, which serves as a potential surface for indirect prompt injection.
Ingestion points: Reads documents, image metadata, and link structures from a user-specified knowledge base root (SKILL.md).
Boundary markers: The instructions do not define specific delimiters (such as XML tags) for isolating processed content from agent instructions.
Capability inventory: The skill is authorized to perform file system operations (read, write, move, archive) and utilize host-provided CLI and browser capabilities.
Sanitization: The accompanying profile_store.py script implements a blocklist for sensitive keywords (e.g., 'token', 'secret', 'api_key') to prevent credentials from being persisted in the user profile.
[COMMAND_EXECUTION]: The workflow involves the agent executing file system commands and local scripts (profile_store.py, validate_skill.py) to manage configuration and perform the organization tasks.
[SAFE]: The provided Python utilities adhere to security standards. Specifically, profile_store.py uses atomic writes with restricted file permissions (0o600), and validate_skill.py uses yaml.safe_load() to prevent arbitrary code execution during YAML parsing.

lov-kb-organize