Skip to content

Changelog

0.11.0

Added

  • Added kardbrd card metadata get, set, remove, and update for revision-checked card metadata with arbitrary JSON values and literal keys. Deploy the Django metadata endpoint and additive migration before using this CLI release; older servers reject these calls.

0.10.0

Changed

  • Collection, search, and activity commands now default to stable TSV with headers. Use --no-headers for headerless TSV, --format json for the previous lossless JSON output, or --format md for Markdown tables. (#58)
  • The non-root agent image now includes Go, GitHub CLI, and pre-commit, and pins Codex CLI 0.144.5. (#59)

Added

  • Added kardbrd self-update, which selects the current platform's GitHub release archive, verifies its SHA-256 checksum, validates its contents, and atomically replaces the installed executable. (#60)
  • Added release archives for Linux and macOS on amd64 and arm64. (#60)
  • Added kardbrd card update --clear-labels for removing every label from a card. (#55)

Fixed

  • Fixed card update --label and --label-ids so they validate, de-duplicate, and reconcile a card's complete label set. (#55)
  • Fixed board labels JSON and Markdown output to use the board's label catalog. (#56)
  • Restored the agent runtime's ability to create pull requests, inspect CI, and push workflow-file changes. (#59)

0.1.0

Added

  • Bot resilience: catch unhandled exceptions that kill WebSocket loop (#40)
  • Retry-via-resume for recoverable Claude exit code failures (#39)
  • Diagnostic error reporting and Claude log extraction (#38)
  • Resolve full CLI path with shutil.which before spawning executors (#37)
  • CodexExecutor for OpenAI Codex CLI integration (#36)
  • Watchtower and uv deployment examples (#35)
  • Replace MCP tool references with kardbrd CLI commands in skills (#33)
  • Migrate commands to skills directory format (#32)
  • Replace MCP server with kardbrd CLI for board operations
  • SOUL.md, RULES.md, and SKILL.md frontmatter for GitAgent standards
  • __self__ support for assignee condition in rule engine
  • comment_author condition with __self__ support (#30)
  • assignee condition with strict YAML list validation (#27)
  • On-demand WebSocket streaming for agent subprocess output (#28)
  • Register bot skills via PUT /api/bots/skills/ on startup (#29)
  • kardbrd API token validation on startup (#23)
  • Skip worktree management when cwd is not a git repository (#22)
  • Enhanced bot card with executor, auth, skills, schedules, and version info (#25)
  • Fix [Errno 7] Argument list too long by piping prompts via stdin (#24)
  • Cron schedule support for time-based automation (#21)
  • Bot card creation/update on startup with triggers display (#20)
  • Wizard card auto-creation on startup when bot is unconfigured (#19)
  • Executor abstraction for multi-agent support — Claude + Goose (#18)
  • Claude CLI auth check before spawning sessions (#17)
  • Deterministic stop confirmation when agent is stopped via reaction (#16)
  • Environment variable support for start command (#15)
  • Top-level config in kardbrd.yml, eliminate statefile (#14)
  • require_label/require_user conditions and Agent-scoped workflows (#13)
  • emoji condition for reaction-based triggers (#11)
  • yamllint and kardbrd.yml validation to pre-commit hooks (#12)
  • kardbrd.yml validation tool with CLI command (#10)
  • exclude_label condition for rule engine (#9)
  • kardbrd.yml rule engine for declarative board automation (#8)
  • --strict-mcp-config to prevent agent using wrong MCP server (#7)
  • Labels support: prompt context and session tracking (#6)
  • Replace FastMCP in-process proxy with kardbrd-mcp stdio subprocess
  • Reaction-based stop replacing word-based stop command (#4)
  • Fetch and update main branch before creating worktrees (#1)
  • Initial release: WebSocket listener, worktree management, Claude executor, MCP proxy