The Agent Skills Directory

[CREDENTIALS_UNSAFE]: Scripts within the testing library (tests/lib_container.sh and tests/claude-devcontainer/lib/container.sh) contain logic to extract authentication data from sensitive local sources. Specifically, the function get_oauth_credentials uses the macOS security CLI to retrieve passwords from the Keychain, and get_api_key_from_config reads API keys directly from ~/.claude.json to configure the agent's authentication in Docker containers.
[REMOTE_CODE_EXECUTION]: Multiple Dockerfiles and initialization scripts (tests/Dockerfile.dev-enhanced, tests/enhanced/setup-enhanced.sh, tests/claude-devcontainer/Dockerfile.enhanced) install external tools using the high-risk curl | sh pattern. This includes the installation of the Starship prompt, the Rust toolchain via rustup, and the Docker CLI.
[COMMAND_EXECUTION]: The skill provides a comprehensive set of shell scripts for container orchestration and environment management. These scripts execute broad commands via subprocess and grant the container user passwordless sudo privileges to allow runtime installation of system, Python, and Node.js packages.
[DYNAMIC_EXECUTION]: tests/remediate_tests.py implements an automated feedback loop where an AI model analyzes routing failures and generates structural modifications to the agent's own instruction files (SKILL.md). These changes are automatically applied to the local file system using the SkillEditor utility, representing a dynamic self-modification capability.

jira-assistant