The Agent Skills Directory

[REMOTE_CODE_EXECUTION] (CRITICAL): The skill provides instructions to download and immediately execute shell and PowerShell scripts from a non-trusted GitHub user repository. This bypasses all security verification.
Evidence: curl -sSL https://raw.githubusercontent.com/yoanbernabeu/grepai/main/install.sh | sh in SKILL.md.
Evidence: irm https://raw.githubusercontent.com/yoanbernabeu/grepai/main/install.ps1 | iex in SKILL.md.
[EXTERNAL_DOWNLOADS] (HIGH): The skill recommends installing Ollama using a piped-to-shell method from an external domain (ollama.com) that is not on the trusted whitelist.
Evidence: curl -fsSL https://ollama.com/install.sh | sh in SKILL.md.
[COMMAND_EXECUTION] (MEDIUM): The skill requires the user to run the grepai daemon, which indexes local source code and creates local storage files (.grepai/index.gob), presenting a significant file system access footprint.
[DATA_EXPOSURE] (LOW): The skill's primary function involves indexing local code projects. While intended, this creates a data ingestion surface where sensitive code content is processed and stored locally in a vector index.
[INDIRECT_PROMPT_INJECTION] (LOW): Vulnerability surface identified.
Ingestion points: Local project files read via grepai watch.
Boundary markers: Absent; the skill does not specify how to handle malicious instructions embedded in source code comments.
Capability inventory: grepai watch (file read/index), grepai search (data retrieval to context), grepai trace (code analysis).
Sanitization: Absent; no mention of sanitizing or escaping code content before returning search results to the agent.

grepai-quickstart