The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The file references/docker-compose-for-multi-container.md contains hardcoded credentials within environment variables for a PostgreSQL service. Evidence: The configuration defines DATABASE_URL=postgresql://postgres:password@db:5432/myapp and POSTGRES_PASSWORD: password, which represents a hardcoded secret pattern in a reference guide.
[INDIRECT_PROMPT_INJECTION]: The skill includes a shell script (scripts/validate-pipeline.sh) that accepts a file path as an argument, creating a surface for indirect prompt injection if the agent processes untrusted pipeline files. Ingestion points: Command-line argument to scripts/validate-pipeline.sh. Boundary markers: No delimiters or instructions are present to mitigate injection from the processed file. Capability inventory: The script currently only executes echo commands; it lacks network or dangerous system call capabilities. Sanitization: No input validation is performed on the provided file path or its contents.

docker-containerization