The Agent Skills Directory

[CREDENTIALS_SAFE]: The skill explicitly warns against hardcoding API keys and provides examples of how to safely load them from environment variables (e.g., process.env.QDRANT_API_KEY). It includes a 'Bad Example' specifically to educate users on why hardcoding credentials in version control is a risk.
[INDIRECT_PROMPT_INJECTION]: As a tool for managing vector embeddings and payloads, the skill inherently possesses an indirect prompt injection attack surface, as it processes and retrieves data that may originate from untrusted external sources.
Ingestion points: Data is ingested into the system via upsert and query operations shown in examples/core.md and examples/recommendations-batch.md.
Boundary markers: No specific delimiters or boundary markers are suggested for the payload data strings.
Capability inventory: The skill facilitates network communication with Qdrant clusters and performing write/read operations on the stored data.
Sanitization: While the skill documentation focuses on technical database configurations (like indexing and dimensions), it does not provide specific sanitization logic for the content of stored strings, which is typical for a database client library.
[EXTERNAL_DOWNLOADS]: The restoreCollection feature in examples/recommendations-batch.md utilizes the Qdrant API's ability to recover data from a remote snapshot URL. This is a standard administrative capability for database management and is documented neutrally as a functional feature.

api-vector-db-qdrant