The Agent Skills Directory

[DATA_EXPOSURE_AND_EXFILTRATION]: The skill accesses the local file system to manage Obsidian notes based on the OBSIDIAN_VAULT environment variable. It does not contain any hardcoded credentials or network communication logic, ensuring that all data processing remains local.
[REMOTE_CODE_EXECUTION]: The scripts/capture.py utility relies exclusively on the Python standard library. It does not download external packages, use dynamic code execution functions like eval() or exec(), or invoke external processes.
[COMMAND_EXECUTION]: The skill does not utilize subprocess, os.system, or any other methods to execute shell commands. All operations are performed using standard file I/O and string manipulation.
[INDIRECT_PROMPT_INJECTION]: The skill facilitates the reading and writing of user-generated content in Markdown format. While it does not implement explicit sanitization for instructions embedded within note text, the attack surface is limited to Markdown modifications within the designated vault directory.
Ingestion points: User-provided CLI arguments and existing content from .md files in the DailyPlan and Weekly directories.
Boundary markers: The script uses standard Markdown task (- [ ]) and bullet point syntax as delimiters.
Capability inventory: File reading and writing via the pathlib module. No network or administrative capabilities are present.
Sanitization: Content is written directly to files without escaping or filtering.

obsidian-capture