The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to clone a repository from an external, personal GitHub account (https://github.com/bikramkgupta/do-app-devcontainer.git) to obtain devcontainer templates and scripts.
[REMOTE_CODE_EXECUTION]: Scripts downloaded from the external repository, such as init.sh, post-create.sh, and various test scripts in the tests/ directory, are executed on the host or inside containers during the provisioning and verification workflows.
[COMMAND_EXECUTION]: The skill provides instructions for executing shell commands with elevated privileges (e.g., sudo chown -R vscode:vscode /home/vscode) and running arbitrary scripts inside containers via docker exec.
[CREDENTIALS_UNSAFE]: Multiple connection strings and configuration templates contain hardcoded default passwords for database services (e.g., postgresql://postgres:password@postgres:5432/app, mysql://mysql:mysql@mysql:3306/app, rustfsadmin:rustfsadmin). While common for development, these represent a security best-practice violation.
[PROMPT_INJECTION]: The skill presents an indirect prompt injection surface in Workflow 2, where it reads and processes untrusted data from .do/app.yaml to automate local container profile mapping.
Ingestion points: Reads .do/app.yaml to detect production service configurations (Workflow 2, SKILL.md).
Boundary markers: Absent; the agent is instructed to map engines directly to local profiles without sanitization or boundary delimiters.
Capability inventory: The agent has capabilities to perform git clone, docker compose up, docker exec, and npm install based on the processed configuration.
Sanitization: No validation or sanitization of the values found in .do/app.yaml is performed before they are used to determine tool execution parameters.

devcontainers