The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill describes Retrieval-Augmented Generation (RAG) workflows that ingest untrusted data from external vector storage.\n
Ingestion points: Data is retrieved via client.search and client.scroll in SKILL.md and references/advanced-usage.md.\n
Boundary markers: The examples do not demonstrate the use of delimiters to isolate retrieved context from the agent's system instructions.\n
Capability inventory: The skill relies on qdrant-client for data retrieval and standard LLM prompt construction. No high-risk capabilities (like shell execution) are applied to the ingested data.\n
Sanitization: The documentation does not specify sanitization or validation steps for payloads fetched from the database.\n- [EXTERNAL_DOWNLOADS]: The documentation references official packages and Docker images from well-known sources.\n
Recommends installing qdrant-client from standard registries.\n
Mentions pulling official images from the qdrant/qdrant Docker Hub repository.\n- [PRIVILEGE_ESCALATION]: The troubleshooting guide includes commands requiring elevated permissions.\n
references/troubleshooting.md suggests sudo systemctl start docker to resolve environment issues.\n
The use of sudo is confined to illustrative troubleshooting text for the user.

qdrant-vector-search