The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted content from external RSS/Atom feeds provided by the user. This data is interpolated into the agent's context, which could allow for indirect prompt injection if a feed contains malicious instructions.
Ingestion points: Content is fetched from the user-provided feed_url in SKILL.md.
Boundary markers: No specific boundary markers or delimiters are defined to separate feed content from agent instructions.
Capability inventory: The skill invokes remote AgentPMT-hosted tools; no local file system, network, or subprocess capabilities are present in the provided files.
Sanitization: No content sanitization or instruction-filtering is implemented for the retrieved feed data.
[EXTERNAL_DOWNLOADS]: The documentation provides instructions to install related setup skills using the skills.sh or openclaw managers from the vendor's own repository (AgentPMT/agent-skills).

rss-reader