The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides fallback instructions to execute a shell command if the primary MCP tool fails. It attempts to locate a local JavaScript bundle (cli.bundle.mjs or cli.js) relative to the skill's installation directory and execute it using node.\n- [PROMPT_INJECTION]: The skill instructs the agent to display diagnostic results verbatim from tool or command output. This creates an indirect prompt injection surface where untrusted data enters the conversation context without sanitization or boundary markers.\n
Ingestion points: Output from the ctx_doctor MCP tool or the fallback shell command stdout.\n
Boundary markers: None (instructions specify to display results verbatim).\n
Capability inventory: The skill performs shell execution via node and displays the resulting stdout to the user.\n
Sanitization: None (no validation or escaping of the output is specified).

ctx-doctor