The Agent Skills Directory

[CREDENTIALS_UNSAFE]: Hardcoded authentication credentials for a Piwigo media gallery hosted at media.delo.sh are present in the instruction set.
Evidence: username="delorenj", password="Ittr5eesol" in Phase 6 of SKILL.md.
[COMMAND_EXECUTION]: The skill executes local Python scripts and shell commands for image processing, vectorization, and packaging.
Evidence: Execution of scripts/coloring-convert, magick for image conversion, and zip for packaging in SKILL.md.
[EXTERNAL_DOWNLOADS]: The skill interacts with external services to process images and upload assets.
Evidence: Uploads and downloads assets via fal.ai (a well-known AI service) and media.delo.sh (a vendor-controlled resource).
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted image data from the web which could theoretically contain malicious instructions for the AI models used in the pipeline.
Ingestion points: Captures external image content via Google Images search results and browser screenshots (SKILL.md Phase 1 and 2).
Boundary markers: Absent.
Capability inventory: Spawns sub-agents (sessions_spawn), executes local scripts, performs file writes, and makes network requests.
Sanitization: No sanitization or filtering is applied to the image content before it is processed by the fal-ai/qwen-image-edit-2511 model.

creating-coloring-books