The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides numerous commands for the agent to execute in the user's environment, including filesystem operations (ls node_modules/@mastra/, cat node_modules/...), package management (npm install, npm update), and running development servers (npm run dev). These are standard for a development framework guide but involve broad access to the local project structure.
[EXTERNAL_DOWNLOADS]: The skill instructs the agent to fetch the latest documentation and migration guides from the official Mastra website (https://mastra.ai/llms.txt). These fetches are used to provide the agent with current API signatures and patterns.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) as it ingests untrusted data from external sources and local files to guide its behavior.
Ingestion points: Files in node_modules/@mastra/ (specifically SOURCE_MAP.json and .d.ts files) as described in SKILL.md, and remote content from https://mastra.ai/llms.txt as described in references/remote-docs.md.
Boundary markers: Absent. There are no explicit instructions for the agent to use delimiters or ignore embedded instructions within the fetched documentation.
Capability inventory: The skill uses subprocess calls (cat, ls, npm, npx) and network operations (WebFetch) to process the ingested data.
Sanitization: Absent. The skill does not specify any validation or filtering of the content retrieved from local files or remote URLs before incorporating it into the response context.

mastra