The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill uses the multica skill import --url command to download agent skill definitions from external hosts including GitHub, ClawHub, and Skills.sh.
[COMMAND_EXECUTION]: The skill is authorized to use the multica CLI within a Bash environment to perform workspace management tasks such as skill registration, updating provenance, and binding skills to agents.
[REMOTE_CODE_EXECUTION]: The import functionality allows the agent to fetch and install new instruction sets and supporting scripts from remote repositories, which are then integrated into the workspace execution environment.
[INDIRECT_PROMPT_INJECTION]: The skill provides a surface for ingesting untrusted instructions through external URLs or uploaded archives.
Ingestion points: The --url and --file parameters in the multica skill import command are the primary entry points for external content.
Boundary markers: The skill does not currently implement specific boundary markers for the imported content at the time of ingestion; protection relies on the platform's handling of the imported skill.
Capability inventory: The skill has access to the Bash tool with multica command scope.
Sanitization: The skill documentation notes that the server-side implementation validates archives against path traversal (zip-slip) and enforces specific file size, count, and bundle limits.

multica-skill-importing