The Agent Skills Directory

[DATA_EXPOSURE]: The skill reads a local file at ~/.uplika/<workspace-slug>/voice.md to personalize content tone. This access is limited to the vendor's own application directory and does not target general system secrets or sensitive credentials.
[COMMAND_EXECUTION]: The instructions include a step to use curl for uploading media bytes to an uploadUrl provided by the vendor's API. This is a standard operational procedure for file uploads in this environment.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from the local voice.md file to influence story generation, which presents a potential injection surface if the file contains malicious instructions.
Ingestion points: ~/.uplika/<workspace-slug>/voice.md (referenced in SKILL.md).
Boundary markers: None explicitly defined for the file read; however, the skill mandates that the user approves all frames before publishing.
Capability inventory: File system read, network media upload (media_presign, media_complete), and publishing to Instagram via publish tool.
Sanitization: Output is processed by the ig-human skill and requires final user confirmation before any action is taken.

ig-story