The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The catalog.json file defines an installation command curl -s https://ethskills.com/l2s/SKILL.md that fetches content from an external domain (ethskills.com) during setup.
[PROMPT_INJECTION]: The skill instructions in catalog.json explicitly guide the user to tell the agent to read external content from https://ethskills.com/l2s/SKILL.md. This represents an indirect prompt injection vector where the agent is encouraged to ingest untrusted data from a remote server.
Ingestion points: https://ethskills.com/l2s/SKILL.md (specified in catalog.json setup and install fields).
Boundary markers: None present in the provided files to delineate external content from system instructions.
Capability inventory: The skill itself does not include executable scripts or tool-use capabilities, limiting the potential impact of an injection.
Sanitization: None detected; the skill relies on the agent directly reading the remote URL.

Layer 2s