The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted content from external documentation sites and changelogs, creating a surface for indirect prompt injection.
Ingestion points: Data is retrieved from documentation URLs and GitHub via mcp__Ref__ref_read_url, mcp__context7__query-docs, and WebFetch in SKILL.md.
Boundary markers: The skill does not implement delimiters or specific instructions (e.g., "ignore embedded commands") to wrap the external data being processed.
Capability inventory: The agent's capabilities in this context are limited to technical information retrieval and reporting; it possesses no file system write, shell execution, or network exfiltration capabilities beyond returning research results to the caller.
Sanitization: There is no defined process for sanitizing or validating the fetched text before it is analyzed by the agent.
[EXTERNAL_DOWNLOADS]: The skill retrieves information from external services as a core part of its technical research functionality.
Evidence: Step 3 explicitly references fetching raw files from GitHub repositories (CHANGELOG.md or CHANGES.md).
Context: These downloads target established documentation platforms and well-known repositories associated with software libraries.

api-state