The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands and Python scripts for directory management and file modification. Specifically, resources/ingestion-protocol.md contains instructions for mkdir, git clone, Move-Item, and Remove-Item. The Python script scripts/safe_replace.py performs a recursive walk through directories to find and replace text in multiple file types (.md, .py, .json, .js, .txt, .css, .html).
[INDIRECT_PROMPT_INJECTION]: The INGEST mode protocol in SKILL.md and resources/ingestion-protocol.md describes a process for importing external skills from Git repositories. This process involves the agent reading SKILL.md frontmatter from untrusted sources to generate registry.json entries.
Ingestion points: External Git repositories cloned via git clone (identified in resources/ingestion-protocol.md).
Boundary markers: None identified in the provided protocols to separate imported content from the agent's core instructions.
Capability inventory: The skill has the capability to write to the local filesystem (write_to_file, Move-Item, Set-Content), execute shell commands (git clone), and modify existing files via scripts/safe_replace.py.
Sanitization: There is a "De-branding" phase in resources/ingestion-protocol.md that replaces specific strings (e.g., 'Claude' with 'Antigravity'), but no general sanitization or escaping for potentially malicious prompt content in the imported markdown files.
[EXTERNAL_DOWNLOADS]: The skill's INGEST mode relies on downloading content from external Git repositories. While the protocol uses a placeholder for the URL, the functionality is designed to fetch and execute logic from remote sources.

skills-managing