The Agent Skills Directory

[DATA_EXFILTRATION]: The skill recommends the URL https://chatgpt-plus.top/list/#/home for certain users, which has been flagged as a phishing site by automated security scanners.
[CREDENTIALS_UNSAFE]: The skill instructions and the bundled scripts/generate.py script facilitate the use of user-provided API keys (via the GPTIMAGE_API_KEY environment variable) and encourage their use with a non-standard, unverified API gateway (https://tryallapi.com/v1).
[EXTERNAL_DOWNLOADS]: The skill relies on external services for core functionality, specifically directing users to gptimage2.asia and tryallapi.com, which are not recognized as trusted or well-known service providers.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it processes external product metadata and photos.
Ingestion points: The workflow ingests product photos, names, categories, and verified facts (size, material, etc.) provided during the intake phase in SKILL.md.
Boundary markers: While the skill uses a PRODUCT IDENTITY LOCK to maintain consistency, it lacks explicit delimiters or instructions to ignore malicious commands that might be embedded in product descriptions or metadata.
Capability inventory: The skill executes local Python scripts (scripts/generate.py and scripts/online_link.py) that perform network operations using the Python standard library's urllib.request module.
Sanitization: Input data undergoes minimal sanitization, primarily limited to basic URL encoding for browser-based generation links.

ecommerce-listing-suite