The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest data from external sources (e.g., querying microsoft-docs via MCP) to ensure skill patterns remain current. This creates a surface for indirect prompt injection where malicious documentation could influence the skill creation process.
Ingestion points: MCP tool results from querying Microsoft documentation (SKILL.md).
Boundary markers: None explicitly defined in the processing of external doc queries.
Capability inventory: Shell command execution (ln, cd, pnpm, npx) for skill organization and testing (SKILL.md).
Sanitization: Not explicitly implemented in the guide for handling external documentation content.
[COMMAND_EXECUTION]: The skill instructions include shell commands for the agent to manage the filesystem and run local testing tools.
Evidence: Commands such as ln -s for creating symlinks, pnpm harness for running the test suite, and npx tsx for documentation generation are included as standard parts of the development lifecycle described in SKILL.md.

skill-creator-ms