The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use Bash to open claude:// deep links for session creation and execution of Python snippets for URL encoding.
[DYNAMIC_EXECUTION]: The skill generates and executes local code snippets to perform URL encoding via Python's urllib.parse module to construct session parameters.
[INDIRECT_PROMPT_INJECTION]: The skill establishes a multi-session orchestration workflow where worker sessions communicate through shared handoff files and direct messaging, creating a vector for cross-session instruction propagation.
Ingestion points: Data is ingested through get_session polls, list_events monitoring, and reading shared file system "handoff paths" as described in the SKILL.md instructions.
Boundary markers: The instructions suggest using setup messages such as "Thread setup for . Reply with just: ready", but they lack explicit instructions for sanitizing or delimiting data transferred between session contexts during the execution phase.
Capability inventory: The skill utilizes desktop-specific tools for session management (mcp__ccd_session_mgmt__*) and sidebar manipulation (mcp__ccd_sidebar__*), and has the ability to send arbitrary messages and execute local shell commands.
Sanitization: No filtering, escaping, or verification of the content within handoff files or worker session events is specified in the implementation logic.

native-threads