The Agent Skills Directory

[SAFE]: The skill implements a Python-based parser (scripts/parse_snapshot.py) designed to convert accessibility tree snapshots into Markdown documents. The script's logic is transparent and focuses solely on text manipulation and tree-building.
[COMMAND_EXECUTION]: The Python script processes input and output files using standard file system operations. It does not utilize subprocess, os.system, or any other functions that could lead to arbitrary command execution.
[DATA_EXFILTRATION]: There are no network operations or credential harvesting patterns detected. The script does not make HTTP requests or access sensitive directories.
[PROMPT_INJECTION]: The SKILL.md file contains instructional content regarding tool usage and does not include any patterns intended to bypass AI safety filters or override system instructions.
[REMOTE_CODE_EXECUTION]: No external code is downloaded or executed. The skill relies entirely on the provided local Python script.
[SAFE]: The skill processes data from external web pages (via the accessibility tree), which constitutes a potential surface for indirect prompt injection. However, the script is a dedicated parser with restricted capabilities (text transformation and file writing), and it includes basic boundary markers (e.g., searching for YAML start blocks) which help delineate the data content.

create-md-from-browsermcp-snapshot