v0.4.1

Fixed

• Piped stdin context was silently ignored in interactive mode without a prompt argument (e.g. issue-md <URL> | danger-claude). The context is now passed as the initial prompt to Claude.

v0.4.0

Added

• Stdin piping support: pipe any content into danger-claude as context. In print mode (-p), the piped content is prepended to the prompt. In interactive mode, it is mounted as /context/stdin inside the container. Stdin is reopened from /dev/tty after consumption so interactive Docker sessions retain a working TTY. Example: issue-md <URL> | danger-claude -p "fix this bug".
• Positional prompt argument: danger-claude "explain this code" starts an interactive Claude session with that initial prompt, like the claude CLI itself.
• -s / --shell now accepts an optional command: danger-claude -s "ls -la" runs the command in the container instead of opening a bash shell.
• -P / --port MAPPING flag to expose container ports to the host (passed as -p to docker run). Supports multiple -P flags for multi-port setups.

Changed

• Breaking: bare arguments are now treated as a Claude prompt instead of a Docker command. Use -s "command" to run arbitrary commands in the container (replaces -- command).

v0.3.1

Added

• --add-host=host.docker.internal:host-gateway on Linux so containers can reach host services (e.g. Chrome DevTools).

v0.3.0

Added

• -r / --resume SESSION_ID flag to resume a previous Claude session. Works in both print mode (-p -r) and interactive mode (-r alone).

v0.2.1

Added

• --version CLI flag to display the current version.

v0.2.0

Added

• New -m/--model MODEL option to select the Claude model (forwarded as --model to the claude CLI).
• New -e/--effort LEVEL option to set effort level: low, medium, high (forwarded as --effort to the claude CLI).

Changed

• -c (auto-commit) now instructs Claude to use Conventional Commits format (<type>: <description>) instead of free-form summary lines.

v0.1.1

Added

• New --max-turns N, --output-format FORMAT, and --json-schema SCHEMA options forwarded to the claude CLI in print mode. Enables compatibility with tools like mr-review that rely on these Claude CLI flags.

v0.1.0

Added

• New -a/--agent NAME option to run in print mode with a specific Claude Code subagent. Passes --agent NAME to the claude CLI, enabling subagent-specific system prompts, tool restrictions, models, and persistent memory.

v0.0.6

Added

• Activate mise in the entrypoint and automatically trust all mounted volumes (workdir + extra -v paths) via mise trust so that project .mise.toml / .tool-versions are picked up automatically.

v0.0.5

Added

• New -g/--git-rw option to mount .git as read-write. By default .git is mounted read-only; -c implies --git-rw. Works with all modes (-s, -p, default, --).
• Pre-create mount points for named volumes (mise, gh) in the Dockerfile so they inherit claude ownership instead of being created as root by Docker.

Changed

• Rewrite CLI usage banner to document all modes and the read-only/read-write distinction.

Fixed

• Use -i instead of -ti when stdin is not a TTY, allowing non-interactive use from scripts and subprocesses (e.g. autodev).