The Agent Skills Directory

[DATA_EXFILTRATION]: The skill manages authentication tokens within the ~/.bria/credentials file. These tokens are transmitted to vendor-controlled endpoints at engine.prod.bria-api.com to authorize image processing requests. This file access and network activity are strictly aligned with the skill's stated purpose and use of vendor-owned infrastructure.
[COMMAND_EXECUTION]: The provided helper script uses standard shell utilities including curl, sed, grep, and base64. These tools are used to encode local image files, interact with the REST API, and manage the local file system for downloaded assets.
[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes JSON manifests from the Bria API status endpoint to identify and download image layers. 1. Ingestion points: External data enters through the result.json manifest file at the end of the delayering process. 2. Boundary markers: None are used as the data is consumed by a shell script for file management rather than being interpolated into an LLM prompt. 3. Capability inventory: The skill utilizes curl, mkdir, and mv to download and organize asset files. 4. Sanitization: Filenames extracted from the manifest are sanitized using a character whitelist (A-Za-z0-9._-) to prevent path traversal or other file-system-based injection attacks.

ad-delayer