The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from social media via the list_channel_posts tool and saves it directly into a local voice profile file without sanitization. This file is specifically designed to be read by other writing agents, creating a surface where malicious instructions hidden in a user's previous posts could influence the behavior of other AI agents in the same workspace.
Ingestion points: External content fetched via list_channel_posts in SKILL.md.
Boundary markers: Absent; the content is placed under a '## Examples' header but lacks explicit delimiters or instructions to ignore embedded commands.
Capability inventory: The skill has the capability to write to the filesystem at ~/.uplika/<workspace-slug>/voice.md.
Sanitization: No sanitization, filtering, or escaping is performed on the content fetched from the external tool.
[COMMAND_EXECUTION]: The skill instructions direct the agent to interact with the local filesystem (~/.uplika/) to create directories, read files, and save content. While the path is scoped to a vendor-specific subdirectory, this represents a persistent modification of the user's environment based on external and user-supplied data.

voice-profile