The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from an external Nextcloud server, which constitutes an untrusted ingestion point for the agent.
Ingestion points: API responses from PROPFIND, OCS JSON/XML, and file metadata retrieved via curl in SKILL.md and references/api-reference.md.
Boundary markers: The skill includes explicit security guardrails instructing the agent to treat server-provided strings (filenames, user IDs, etc.) as data rather than instructions.
Capability inventory: Network access via curl and local file management via curl -o/-T and split.
Sanitization: The skill provides specific utility functions (nc_urlencode, nc_urlencode_path) using jq to properly escape server-side data before it is interpolated into shell commands.
[COMMAND_EXECUTION]: The skill relies on the execution of local shell commands to perform its functions.
Evidence: Use of curl, jq, uuidgen, and split in SKILL.md and references/api-reference.md for API interaction, data parsing, and handling large file uploads.

nextcloud-admin