The Agent Skills Directory

Command Execution: The skill uses the Bash tool to interact with the ms command-line utility for managing share links. These operations are gated by a check for the project configuration file (ms.config.json) and a verification of the user's authentication status. Furthermore, the skill instructions require the agent to obtain manual confirmation from the user before revoking access links, ensuring that sensitive operations are not performed autonomously without oversight.
Indirect Prompt Injection Surface: The skill processes information from the local project environment which acts as a data ingestion point. (1) Ingestion points: The agent reads the contents of ms.config.json, memory-bank.md, and the output generated by CLI commands like ms app share link list. (2) Boundary markers: The instructions do not define specific delimiters or headers to isolate external data from the system prompt. (3) Capability inventory: The skill has access to tools for command execution (Bash) and file system interaction (Read, Grep, Glob). (4) Sanitization: The skill does not explicitly perform sanitization or validation of the data retrieved from configuration files or command outputs before it is interpreted by the agent. While this represents a potential surface for indirect instructions, it is standard for skills operating within a trusted development workspace.

share-link