The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill contains a hardcoded password string ('securePassword123') within a TypeScript code example in SKILL.md under the "Sharing & Permissions" section. While presented as an example, the use of a non-placeholder secret value in instructions is a security risk.
[COMMAND_EXECUTION]: The instructions utilize the Node.js fs module to perform local filesystem operations, including fs.readFileSync, fs.createWriteStream, fs.openSync, and fs.readSync. These are used to handle file uploads and downloads, granting the agent the capability to interact with the host filesystem.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from an external cloud service (Microsoft OneDrive), which could contain malicious instructions embedded in file names or content.
Ingestion points: Data retrieved from OneDrive API endpoints such as /drive/root/children and /drive/items/{id}/content in SKILL.md.
Boundary markers: The instructions lack explicit boundary markers or delimiters to separate external data from agent instructions.
Capability inventory: The skill has the ability to read and write to the local filesystem (fs module) and perform authenticated network operations via the graphClient.
Sanitization: There are no procedures defined for validating or sanitizing the data fetched from the external API before processing.

onedrive