The Agent Skills Directory

CREDENTIALS_UNSAFE (HIGH): The skill contains a hardcoded secret token in multiple example curl commands.
Evidence: secret=tsh-revalidate-2024 is included in the revalidation URLs provided for cache management.
Impact: An attacker could use this secret to trigger unauthorized cache revalidations on the tsh.sale production environment, potentially leading to denial-of-service or data consistency issues.
PROMPT_INJECTION (LOW): The skill exposes a surface for Indirect Prompt Injection (Category 8) by ingesting data from external Zoho API endpoints.
Ingestion points: Data is ingested via zohoFetch from Zoho Inventory (/inventory/v1/items) and Zoho Books (/contacts, /salesorders, etc.) endpoints.
Boundary markers: None identified. The documentation does not specify the use of delimiters or instructions to ignore embedded content when processing API responses.
Capability inventory: The skill facilitates data fetching, caching via Upstash Redis/Next.js unstable_cache, and serving data through API routes (NextResponse.json). It does not appear to have direct shell execution or file-write capabilities on the host.
Sanitization: There is no evidence of sanitization or filtering for natural language instructions that might be embedded in product descriptions, customer names, or order notes retrieved from Zoho.

zoho-api