The Agent Skills Directory

[EXTERNAL_DOWNLOADS] (MEDIUM): The skill recommends using npx -y @context7/mcp-server. This command downloads and executes code from the npm registry at runtime without specific version pinning or integrity verification, which is a vector for remote code execution if the package or registry is compromised.
[COMMAND_EXECUTION] (LOW): The core workflow relies on the execution of shell commands (curl, jq). While intended for documentation retrieval, it establishes a pattern of the agent interacting directly with the system shell.
[DATA_EXFILTRATION] (SAFE): Network activity is restricted to context7.com. No access to sensitive local files or credentials was detected in the provided scripts.
[PROMPT_INJECTION] (LOW): The skill processes content fetched from an external API (context7.com).
Ingestion points: Data enters the context via curl responses.
Boundary markers: None provided in the examples to distinguish between documentation and potential instructions.
Capability inventory: curl, jq, npx.
Sanitization: No evidence of sanitization or filtering of the fetched documentation before presenting it to the agent.

context7