The Agent Skills Directory

[PROMPT_INJECTION]: Indirect Prompt Injection Surface.
Ingestion points: The skill utilizes the mcp__context7__get-library-docs tool to fetch external documentation at runtime, which is then processed to influence configuration architecture.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to disregard potentially malicious instructions embedded within the fetched documentation.
Capability inventory: The agent is granted high-privilege capabilities including Bash, Write, Edit, and WebFetch, which could be abused if malicious instructions are successfully injected.
Sanitization: No sanitization or validation logic is shown for the content retrieved via the MCP tools.
[COMMAND_EXECUTION]: High-Privilege Tool Access. The skill requests the Bash tool to perform configuration tasks. While this is consistent with an enterprise configuration management role, it grants the agent the ability to execute arbitrary commands on the host system.
[EXTERNAL_DOWNLOADS]: Network Communication and Content Retrieval. The skill leverages WebFetch and specialized MCP tools to download external data. Additionally, code examples illustrate using the fetch API to interact with internal and external services such as HashiCorp Vault and Kubernetes APIs.
[REMOTE_CODE_EXECUTION]: Dynamic Loading from Computed Paths. The TypeScript configuration manager example uses require(configPath) where the path is dynamically constructed from an environment variable. This pattern can be exploited to load unintended local files if the input is not strictly controlled.

moai-cc-configuration