The Agent Skills Directory

[SAFE]: No malicious patterns detected. The skill performs localized file management tasks consistent with its stated purpose of organizing Obsidian notes.
[COMMAND_EXECUTION]: The agent is instructed to execute a bundled Python script (generate_metadata.py) to apply and validate metadata changes. This script handles logic for YAML manipulation and schema validation, representing a secure way to manage file updates compared to raw shell commands.
[DATA_EXPOSURE]: The skill accesses local markdown files to extract content and write frontmatter. There is no evidence of network operations, hardcoded credentials, or exfiltration to external domains.
[INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted data from raw transcriptions to generate metadata.
Ingestion points: Raw markdown transcriptions are read from the user's filesystem (SKILL.md).
Boundary markers: None explicitly defined in instructions, though the agent is instructed to distinguish between the body and existing metadata.
Capability inventory: The skill has file-read and file-write capabilities via the generate_metadata.py script (scripts/generate_metadata.py).
Sanitization: The script uses json.dumps to escape YAML scalars and validates all generated metadata against a local schema.json to ensure type and content safety.

generate-metadata