The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest session history (including messages, code changes, and tool outputs) from 16 different AI tools and re-inject that content as a 'handoff document' into a new assistant. This creates a vulnerability where malicious instructions contained in a prior session could influence the behavior of the next assistant.\n
Ingestion points: Session files located in application-specific directories such as ~/.claude/projects/, ~/.cursor/projects/, ~/.copilot/session-state/, and VS Code global storage folders.\n
Capability inventory: The continues CLI tool reads session databases and JSONL files, generates formatted markdown prompts, and is intended to be used as the starting point for new AI assistant sessions.\n
Boundary markers: The generated handoff document uses markdown headers for structure, but the skill does not specify the use of clear delimiters or instructions to the destination AI to ignore potentially malicious embedded content within the history.\n
Sanitization: No sanitization or filtering of the original message content is mentioned before it is passed to the new assistant.\n- [COMMAND_EXECUTION]: The instructions rely on the user executing the continues CLI tool. The tool performs extensive filesystem operations, including scanning sensitive application data directories in the user's home folder and project-specific directories to extract session metadata and content.\n- [EXTERNAL_DOWNLOADS]: The skill recommends running the continues package using npx or installing it globally via npm install. This involves downloading and executing code from an external registry to perform session discovery and extraction tasks.

continues-cli-session-transfer