The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes data from external Ozon search results, keyword lists, and product details via a third-party API. While this is the intended function, it creates an attack surface where malicious search data could theoretically influence the agent's behavior.
Ingestion points: Data is retrieved from the https://www.iqinghu.com/api/desktop/qh/mcp endpoint and specific tool methods such as queryhotkeywordrankings and querykeywordrelatedproducts (found in SKILL.md).
Boundary markers: The instructions do not define explicit boundary markers (e.g., XML tags or delimiters) to separate external data from the system prompt, though it does mandate structured parsing of the JSON content.
Capability inventory: The skill uses network request capabilities (POST/GET) and processes tool outputs across multiple steps.
Sanitization: There are no explicit instructions to sanitize or escape external Russian/English text content before the agent generates advice or summaries.
[CREDENTIALS_UNSAFE]: The skill is configured to read sensitive authentication tokens (QINGHU_TOKEN, QHKIT_TOKEN) from the environment. While using environment variables is a recommended practice for secret management, it identifies the specific keys an attacker might target for environment variable exfiltration. The skill also includes a referral-coded registration link (urlCode=1788417527636) for the third-party service.

qinghu-ozon-keyword-picker