The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes data from the Snapchat Marketing API, which can contain content from external sources that might influence agent behavior.
Ingestion points: External data enters the context through list and get tools such as snapchat_ads_list_campaigns, snapchat_ads_list_creatives, and snapchat_ads_list_media (found in references/discovery.md).
Boundary markers: There are no instructions for the agent to use delimiters or ignore embedded instructions within the API responses.
Capability inventory: The skill has significant write and delete capabilities, including snapchat_ads_create_campaign, snapchat_ads_update_ad_squad, snapchat_ads_delete_ad, and snapchat_ads_upload_media (found in SKILL.md and references/campaign-creation.md).
Sanitization: No sanitization or validation of the data retrieved from the API is implemented before the agent processes it.
[DATA_EXFILTRATION]: The skill handles potentially sensitive user data and local file paths.
The snapchat_ads_add_segment_users tool transmits user identifiers like emails and phone numbers. While the instructions in SKILL.md and references/targeting.md explicitly mandate that identifiers must be normalized and SHA-256 hashed before transmission, the agent still handles this data flow.
The snapchat_ads_upload_media tool includes a file_path parameter. While SKILL.md warns to "Avoid server-local paths," the presence of this parameter provides a surface for the agent to access and transmit local files to the Snapchat API.
[COMMAND_EXECUTION]: The toolkit interface for snapchat_ads_upload_media allows the agent to specify file paths for upload, which is a functional command execution surface for file system interaction.

snapchat-ads