The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection because it retrieves and summarizes content derived from external sources like error logs, GitHub issues, and customer support tickets.\n
Ingestion points: Untrusted data enters the context through inbox-reports-list and inbox-reports-retrieve tools, as well as the mentioned signals skill tools.\n
Boundary markers: The instructions do not provide explicit delimiters or warnings to separate external content from the agent's instructions, increasing the risk of the agent following embedded commands.\n
Capability inventory: The skill's capabilities are restricted to read-only API calls and database queries. It does not include tools for file modification, network exfiltration to third-party domains, or subprocess execution.\n
Sanitization: There is no description of sanitization or validation performed on the retrieved report text before it is summarized for the user.\n- [SAFE]: All network references and deep-links point to official 'posthog.com' domains, which align with the skill's stated vendor and purpose.\n- [SAFE]: The skill exclusively employs read-only tools and explicitly states that integration credentials are not accessible through the provided configuration data.

inbox-exploration