The Agent Skills Directory

[SAFE]: No malicious patterns, obfuscation, or unauthorized data access behaviors were detected. The skill's functionality is consistent with its stated purpose of managing SupSub content streams.
[COMMAND_EXECUTION]: The skill executes commands via the supsub CLI for legitimate management tasks. The instructions explicitly direct the AI to perform a 'double confirmation' check with the user before executing destructive commands such as focus remove or focus mark-read --all.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of @supsub/cli. As this package belongs to the official vendor namespace (@supsub) and is the primary tool for the service described, it is categorized as a safe vendor-provided dependency.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external content streams (articles and summaries) via the supsub focus contents command.
Ingestion points: Data is received in JSON format from the CLI tool's output.
Boundary markers: The instructions do not specify explicit delimiters for external content, but provide clear logic for how the agent should interpret the fields.
Capability inventory: The skill has access to focus deletion and status modification commands via the CLI.
Sanitization: No explicit sanitization of article summaries is mentioned, however, the risk is mitigated by the constrained nature of the CLI tool's capabilities and the requirement for user confirmation on sensitive actions.

supsub-focus