The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process Cursor log files and reproduction bundles, which are untrusted external data sources.
Ingestion points: The agent is directed to read Cursor log files from system-specific application support directories (~/Library/Application Support/Cursor/logs/ and ~/.config/Cursor/logs/) and user-provided reproduction snippets.
Boundary markers: Absent. The skill instructions do not require the use of specific delimiters or safety preambles to isolate external content from the agent's primary instructions.
Capability inventory: The skill is granted access to Bash, Write, and Edit tools, creating a functional path for code execution or file modification if an injection occurs within the processed data.
Sanitization: While the user is advised to manually redact sensitive information, there is no programmatic sanitization or validation of the input content by the agent.

cursor-debug-bundle