The Agent Skills Directory

[SAFE]: The skill is designed for social media auditing and operates within the context of its stated purpose. It processes Instagram insights and post metadata without any suspicious execution patterns.
[INDIRECT_PROMPT_INJECTION]: The skill has a data ingestion surface as it reads post captions via the list_channel_posts tool and accepts user-provided text or screenshots. There are no explicit boundary markers or sanitization steps mentioned. However, the available tools (list_accounts, list_channel_posts, get_insights) are restricted to read operations, and the skill lacks the capability to perform network exfiltration or system-level modifications, mitigating the risk of this attack vector.
[DATA_EXPOSURE_AND_EXFILTRATION]: The skill accesses Instagram account IDs and post insights. These operations are conducted through the developer's own infrastructure (Uplika) and are necessary for the skill's functionality. No sensitive environment variables, credentials, or local system files are accessed.

ig-audit