The Agent Skills Directory

[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it scrapes metadata from untrusted external websites and provides it to the agent for analysis without sanitization or boundary markers.\n
Ingestion points: scripts/generate-visuals.py fetches data from user-supplied URLs and extracts metadata like og:title.\n
Boundary markers: None identified; the agent reads the scraped products.json manifest directly.\n
Capability inventory: The skill possesses network access, file write capabilities, and integrates with the Fal.ai image generation API.\n
Sanitization: While filenames are slugified, the product titles and descriptions are processed by the agent in their raw form.\n- [DATA_EXFILTRATION]: The scraper in scripts/generate-visuals.py performs network requests to arbitrary URLs without validating the target addresses, creating a potential Server-Side Request Forgery (SSRF) surface if directed at internal or private network resources.\n- [DATA_EXFILTRATION]: The environment loader in scripts/generate-visuals.py recursively searches for and parses .env files up to five levels above the current directory, which could inadvertently load and expose sensitive credentials from parent project folders.

product-visual-generator