The Agent Skills Directory

[SAFE]: The skill uses a defined set of tools (search, query, get_page, list_pages, takes_search, find_contradictions, find_trajectory) which are scoped to searching and reading content from the platform's knowledge base. No tools for file system modification, network access, or privilege escalation are utilized.
[SAFE]: The skill instructions explicitly set mutating: false in the metadata, ensuring that it remains in a read-only state unless a user explicitly requests a saved artifact, which then routes to other operations. This adheres to the principle of least privilege.
[SAFE]: No evidence of prompt injection, obfuscated code, or hidden malicious instructions was found in the documentation or the routing evaluation fixtures. The logic follows a standard multi-phase approach for information retrieval and classification.
[SAFE]: The skill does not reference or download any external dependencies, scripts, or remote content. All processing is focused on the user's own data context.

idea-lineage