The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to fetch and analyze content from external, user-provided website URLs to identify cookies and tracking technologies, which constitutes an ingestion point for untrusted data.
Ingestion points: Step 2 of the workflow in SKILL.md (and SKILL.fr.md) directs the agent to visit a target URL provided by the user to perform research.
Boundary markers: The instructions do not specify the use of delimiters or specific safety warnings when processing this external content.
Capability inventory: The agent utilizes WebFetch to read external sites and subsequently performs text generation to draft a policy.
Sanitization: No specific sanitization or escaping of the fetched content is mentioned before it is processed by the agent.
Mitigation: The risk is partially mitigated by the mandatory requirement in Step 1 and Step 3 to use a pre-defined, validated template (assets/sample_template_politique_cookies.docx), which restricts the agent from drafting the policy from scratch or deviating from established wording.
[EXTERNAL_DOWNLOADS]: The skill retrieves current regulatory guidelines and FAQs from official government resources using the WebFetch tool.
Evidence: Fetches data from multiple official URLs on cnil.fr, such as https://www.cnil.fr/fr/cookies-et-autres-traceurs/regles/cookies.
Context: These references target a well-known official organization (the French Data Protection Authority) and are essential for providing accurate and up-to-date legal guidance.

cookie-policy-malik-taiar