The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to fetch and process content from untrusted external skill registries and GitHub repositories.
Ingestion points: External SKILL.md files, repository descriptions, and metadata are ingested into the agent's context during the discovery and inspection phases in SKILL.md and references/upstream-skill.md.
Boundary markers: There are no explicit structural delimiters or instructions to ignore embedded commands within the fetched external content, increasing the risk that a malicious skill could hijack the agent's session.
Capability inventory: The agent is instructed to read the full body of top candidates to analyze their scripts, hooks, and permissions, which provides a direct vector for instructions within those files to execute via the agent's reasoning.
Sanitization: The instructions do not specify any sanitization, filtering, or validation of the external markdown content before it is processed.
[COMMAND_EXECUTION]: The skill uses local shell commands to execute python scripts/search.py for performing semantic searches against a local or remote index.
[EXTERNAL_DOWNLOADS]: The skill fallback logic involves performing web and GitHub searches to discover and fetch code (skills) from unknown third-party authors and registries when local discovery fails.

skillfinder-external-scout