The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on shell command execution (curl) to interact with the Nextcloud OCS API and WebDAV interface for all operations including creating, reading, and deleting wiki pages.
[INDIRECT_PROMPT_INJECTION]: The skill ingests markdown content from an external Nextcloud instance. This data is untrusted and could be manipulated by an attacker to include hidden instructions that influence the agent's behavior during content processing or editing tasks.
Ingestion points: SKILL.md and references/webdav-content.md describe fetching page content and metadata via curl GET requests to a remote server.
Boundary markers: The provided examples do not demonstrate the use of delimiters or 'ignore' instructions when presenting retrieved content to the AI agent.
Capability inventory: The skill possesses the ability to perform network operations (curl), write to the local file system (printf, sed), and manipulate the Nextcloud wiki structure via OCS API calls.
Sanitization: While path-based URL encoding is used for safety, the actual content of the markdown files is processed without sanitization.
[DYNAMIC_EXECUTION]: The skill provides fallback logic for JSON parsing using python3 -c to execute a small Python script provided as a string literal. While the script is static and based on a template, it represents runtime script execution based on external JSON input.

nextcloud-collectives