The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill facilitates fetching external content via an unverified third-party service (defuddle.md).
Evidence: The instructions include a command curl -sL "https://defuddle.md/<url>" to retrieve markdown versions of external advisories and specifications.
[DATA_EXFILTRATION]: Using the defuddle.md service involves sending the URL being researched to a third-party server. While the skill includes a warning against routing sensitive or adversary infrastructure through this service, the mechanism inherently shares metadata (the URLs of interest) with an external entity.
Evidence: Found in the 'Reading External Sources' section of SKILL.md.
[PROMPT_INJECTION]: The skill instructs the agent to ingest and process content from external URLs. This content is untrusted and could contain malicious instructions designed to influence the agent's behavior (Indirect Prompt Injection).
Ingestion points: External URLs fetched via curl through the defuddle.md service.
Boundary markers: The skill does not provide specific delimiters or instructions for the agent to ignore embedded commands within the fetched content.
Capability inventory: The agent has the capability to perform network operations (curl) and potentially access local environment configurations during cloud hardening tasks.
Sanitization: No sanitization or validation of the fetched markdown content is described.

hardening-cloud-posture