The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructions provide a command to install the CLI utility using curl -fsSL https://wzrrd.sh/install.sh | bash. This executes code from a remote server directly in the user's shell, which is a significant security risk as the script is unverified and runs with the user's privileges.
[EXTERNAL_DOWNLOADS]: The skill references and fetches resources from the wzrrd.sh domain, including an installation script and supplemental agent instructions located at /.well-known/agent-skills/.
[COMMAND_EXECUTION]: The skill relies on several local command-line tools (wzrrd and joelclaw) to manage video pipelines, which involves executing subprocesses in the local environment.
[PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection through its handling of external data.
Ingestion points: Untrusted data from the cloud-native pipeline, such as transcripts and workflow status metadata, enters the agent context through video/publish.completed events and status command outputs.
Boundary markers: Absent. There are no specific delimiters or instructions to ignore embedded commands within the processed data.
Capability inventory: The skill has the ability to execute local commands and access specific project directories.
Sanitization: There is no evidence of validation or sanitization for the content returned from the remote wzrrd service.

wzrrd-video