The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The documentation for setting up CI/CD workflows includes an automated installation step that fetches and executes an installation script from Ollama's official domain (ollama.com). This is a standard and expected method for environment configuration within automated testing pipelines.
[COMMAND_EXECUTION]: The skill provides several operational shell scripts (setup-rag-database.sh, seed-documents.sh, check-rag-status.sh) used for infrastructure management. These scripts execute standard commands for database setup, document processing, and health monitoring of local services.
[PROMPT_INJECTION]: The skill implements Retrieval-Augmented Generation patterns which represent an indirect prompt injection surface. This occurs when external data is retrieved and interpolated into language model prompts.
Ingestion points: External data enters the agent context through ORM retrieval in references/rag-pipeline.md and file-based seeding in scripts/seed-documents.sh.
Boundary markers: While prompt templates in references/agentic-kit.md use structural labels (e.g., "Context:"), they do not include explicit instructions for the model to disregard potentially adversarial content within the ingested data.
Capability inventory: The skill possesses the capability to execute shell commands (psql, docker, curl) and perform network requests to LLM providers and local APIs.
Sanitization: The routeQuery example in references/agentic-kit.md implements basic regex validation on model output. However, document content is generally concatenated into generation prompts without extensive sanitization or escaping.

constructive-ai