The Agent Skills Directory

[PROMPT_INJECTION]: The skill presents an indirect prompt injection surface through its inbox management capabilities.
Ingestion points: untrusted data enters the agent context via src/commands/inbox.ts through commands that fetch messages, comments, and reviews from external social media platforms.
Boundary markers: absent; the tool outputs raw JSON content from these external sources without delimiters or instructions to ignore embedded commands.
Capability inventory: across the codebase, the tool possesses capabilities to read and write to the filesystem (src/utils/config.ts, src/commands/media.ts, src/commands/contacts.ts, src/commands/sequences.ts) and perform network operations (src/commands/auth.ts, src/commands/media.ts).
Sanitization: absent; external social media content is retrieved and passed directly into the output stream for processing by the agent.
[SAFE]: The skill accesses ~/.zernio/config.json to store and retrieve authentication credentials. This is standard behavior for a CLI tool managing its own configuration.
[SAFE]: The skill performs network requests to zernio.com and vendor-provided presigned URLs for media uploads. These operations are consistent with the tool's primary purpose and target the vendor's own infrastructure.

zernio