The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill downloads asset bundles (.aws/ZIP) from the vendor's domain to update local design presets.
Evidence: import_presets_aws.py uses urllib.request to fetch files from aiworkskills.cn or its subdomains.
[COMMAND_EXECUTION]: The skill executes Python scripts to perform file ingestion and archive processing tasks.
Evidence: SKILL.md instructs the agent to run scripts/product_image_ingest.py and scripts/import_presets_aws.py as part of the core workflow.
[INDIRECT_PROMPT_INJECTION]: The skill ingests external configuration and asset data which can influence the agent's behavior or article output during generation.
Ingestion points: ZIP bundle contents (config.yaml, presets) and user image descriptions processed by the ingestion scripts.
Boundary markers: No explicit delimiter or instruction-guarding markers are applied to the ingested content before it enters the context.
Capability inventory: The skill has permissions to write to the file system (including aws.env and .aws-article/ directories) and execute local scripts.
Sanitization: The import_presets_aws.py script contains robust 'Zip Slip' (path traversal) defense in the _safe_extractall function and implements strict hostname validation for HTTPS downloads.

aws-wechat-article-assets