crossby 0.2.4__tar.gz

crossby-0.2.4/.claude/settings.local.json

@@ -0,0 +1,67 @@
+{
+  "permissions": {
+    "allow": [
+      "Bash(git pull:*)",
+      "WebFetch(domain:pypi.org)",
+      "Bash(wade task:*)",
+      "Bash(.venv/bin/python -m pytest tests/ -x -q)",
+      "Bash(.venv/bin/python -m pytest tests/unit/test_sync/test_mcp_writers.py::TestCodexMCPWriter -q)",
+      "Bash(.venv/bin/python -m pytest tests/ -q --ignore=tests/unit/test_sync/test_mcp_writers.py -k \"not Codex\")",
+      "Bash(.venv/bin/python -m pytest tests/ -q -k \"not Codex and not codex\")",
+      "Bash(.venv/bin/python -m pytest tests/unit/test_sync/test_sync_integration.py::TestFullSyncMCP::test_syncs_to_all_five_tools -q)",
+      "Bash(.venv/bin/python -m pytest tests/ -q -k \"not \\(test_syncs_to_all_five_tools or test_env_vars_preserved_across_all_formats\\)\" --ignore=tests/unit/test_sync/test_mcp_writers.py)",
+      "Bash(.venv/bin/python -m ruff check src/ 2>&1)",
+      "Bash(.venv/bin/python -m mypy src/crossby/)",
+      "Bash(.venv/bin/python -m ruff check src/crossby/config/loader.py --fix 2>&1)",
+      "Bash(.venv/bin/python -m ruff check src/crossby/config/loader.py src/crossby/cli/convert.py src/crossby/sync/permissions.py src/crossby/config/detection.py src/crossby/ai_tools/base.py src/crossby/sync/agents.py src/crossby/sync/rules.py 2>&1)",
+      "Bash(.venv/bin/python -m pytest tests/unit/test_sync/test_mcp_writers.py -k \"not Codex\" -q)",
+      "Bash(git status:*)",
+      "Bash(git checkout:*)",
+      "Bash(python3 -m pytest tests/ -x -q)",
+      "Bash(pip3 list:*)",
+      "Bash(uv run:*)",
+      "Bash(git add:*)",
+      "Bash(gh api:*)",
+      "WebSearch",
+      "Bash(python -m pytest tests/unit/test_handoff -x)",
+      "Bash(git stash *)",
+      "Bash(git commit -m ' *)",
+      "Bash(./scripts/test.sh tests/unit/test_models/test_ai_models.py::TestPlanModeArgs -v)",
+      "Bash(find /Users/ivanviragine/Documents/workspace/crossby/tests -name \"*.py\" -exec grep -l \"launch.*plan\\\\|--plan\" {} \\\\;)",
+      "Bash(./scripts/test.sh tests/)",
+      "Bash(python3 *)",
+      "Bash(find tests *)",
+      "Bash(find *)",
+      "Bash(python -m pytest tests/unit/test_sync/test_readers.py -x -q)",
+      "Bash(uv sync *)",
+      "Bash(./scripts/test.sh)",
+      "Bash(./scripts/check.sh)",
+      "Bash(./scripts/check.sh --types)",
+      "Bash(gh pr *)",
+      "Read(//tmp/**)",
+      "Bash(awk '/^diff --git a\\\\/CONTRIBUTING.md/{found=1} found{print} /^diff --git/ && !/CONTRIBUTING/{if \\(found\\) exit}' /tmp/pr43.diff)",
+      "Bash(awk '/^diff --git a\\\\/src\\\\/crossby\\\\/cli\\\\/launch.py/{found=1} found{print} /^diff --git/ && !/launch.py/{if \\(found\\) exit}' /tmp/pr43.diff)",
+      "Bash(rtk git *)",
+      "Bash(rtk gh *)",
+      "Bash(gh repo *)",
+      "Bash(rtk grep *)",
+      "Bash(bash scripts/check.sh)",
+      "Bash(rtk ls *)",
+      "Bash(rtk read *)",
+      "Bash(env)",
+      "Bash(git check-ignore *)",
+      "Bash(git rm *)",
+      "Bash(./scripts/check-all.sh)",
+      "Bash(rm -rf /tmp/crossby-publish-test)",
+      "Bash(/tmp/crossby-publish-test/bin/pip install *)",
+      "Bash(/tmp/crossby-publish-test/bin/crossby --version)",
+      "Bash(/tmp/crossby-publish-test/bin/crossby --help)",
+      "Bash(/tmp/crossby-publish-test/bin/crossby sync *)",
+      "Bash(rm -rf /tmp/publish-skill-test)",
+      "Bash(mkdir -p /tmp/publish-skill-test/.claude)",
+      "Bash(/tmp/crossby-publish-test/bin/crossby init *)",
+      "Bash(/tmp/crossby-publish-test/bin/crossby convert *)",
+      "Bash(uv build *)"
+    ]
+  }
+}

crossby-0.2.4/.claude/skills/.gitignore

@@ -0,0 +1,6 @@
+# Wade-managed session skills — installed per worktree, not committed
+implementation-session/
+review-pr-comments-session/
+plan-session/
+deps/
+task/

crossby-0.2.4/.gitignore

@@ -0,0 +1,26 @@
+__pycache__/
+*.py[cod]
+*$py.class
+*.egg-info/
+dist/
+build/
+*.egg
+.eggs/
+*.so
+.venv/
+venv/
+.env
+.mypy_cache/
+.pytest_cache/
+.ruff_cache/
+.coverage
+htmlcov/
+*.log
+.commit-msg
+.crossby.yml
+KNOWLEDGE.md
+.agents/
+# wade-managed symlinks (skills sync)
+.cursor/skills
+.gemini/
+.github/skills

crossby-0.2.4/.wade/.gitignore

@@ -0,0 +1 @@
+*

crossby-0.2.4/.wade/.wade-managed

@@ -0,0 +1,2 @@
+.wade.yml
+# Managed by wade 0.20.6

crossby-0.2.4/.wade.yml

@@ -0,0 +1,51 @@
+version: 2
+project:
+  main_branch: main
+  issue_label: feature-plan
+  worktrees_dir: ../.worktrees
+  branch_prefix: feat
+  merge_strategy: PR
+ai:
+  default_tool: claude
+  default_model: claude-sonnet-4.6
+  effort: max
+  deps:
+    tool: copilot
+    model: claude-sonnet-4.6
+    mode: headless
+  review_plan:
+    tool: copilot
+    model: claude-sonnet-4.6
+    mode: headless
+    enabled: true
+  review_implementation:
+    tool: copilot
+    model: claude-sonnet-4.6
+    mode: headless
+    enabled: true
+  plan:
+    tool: claude
+    model: claude-opus-4.7
+  review_batch:
+    tool: copilot
+    model: claude-sonnet-4.6
+    mode: interactive
+    enabled: true
+models:
+  claude:
+    easy: claude-sonnet-4.6
+    medium: claude-sonnet-4.6
+    complex: claude-opus-4.7
+    very_complex: claude-opus-4.7
+provider:
+  name: github
+hooks:
+  post_worktree_create: scripts/setup-worktree.sh
+  copy_to_worktree:
+  - .env
+  - KNOWLEDGE.md
+  - .wade.yml
+  - KNOWLEDGE.ratings.yml
+knowledge:
+  enabled: true
+  path: KNOWLEDGE.md

crossby-0.2.4/CONTRIBUTING.md

@@ -0,0 +1,258 @@
+# Contributing to crossby
+
+Thanks for your interest in contributing. This document is the maintainer/developer guide — architecture, conventions, and how to extend crossby safely. If you're looking for usage docs, see [README.md](README.md).
+
+## Development Setup
+
+Requires Python 3.11+ and [uv](https://docs.astral.sh/uv/).
+
+```bash
+git clone https://github.com/ivanviragine/crossby
+cd crossby
+uv sync --extra dev
+```
+
+## Running Checks
+
+| Command                | What it does                                 |
+| ---------------------- | -------------------------------------------- |
+| `./scripts/test.sh`    | Run the full test suite                      |
+| `./scripts/check.sh`   | Lint (ruff) + type check (mypy strict)       |
+| `./scripts/fmt.sh`     | Auto-format with ruff                        |
+| `./scripts/check-all.sh` | Tests + lint + format check + type check  |
+
+Run `./scripts/check-all.sh` before opening a PR.
+
+`tomli-w` is a base dependency (required for Codex MCP sync), so a plain `uv sync` pulls in everything the test suite needs.
+
+## Architecture
+
+### Directory Layout
+
+```
+src/crossby/
+├── cli/          # Typer commands (entry point: cli/main.py:cli_main)
+├── services/    # High-level orchestrators (sync, launch, handoff)
+├── ai_tools/     # Per-tool adapters (Claude, Copilot, Gemini, Codex, …)
+├── sync/         # Sync writers — translate and write config per tool
+├── subagents/    # Subagent format translation (canonical IR + parsers/emitters)
+├── handoff/      # Session readers, summarizer, prompt loader, handoff writer
+├── config/       # .crossby.yml loading and Pydantic models
+├── models/       # Shared data models (AIToolID, capabilities, …)
+├── data/         # Static model catalog and bundled prompt presets
+├── ui/           # Rich/questionary UI components
+└── logging/      # structlog configuration
+```
+
+### Subagent format translation
+
+`src/crossby/subagents/` translates a single subagent definition between
+Claude Code, Cursor, Gemini, Copilot, and Codex.  The architecture is a
+canonical intermediate representation (`SubagentIR`) with one parser and one
+emitter per tool — no pairwise converters.  Tool-specific fields that don't
+generalize live in `SubagentIR.extras` and are only re-emitted when the
+target tool matches the original source.
+
+Lossy translations surface as `ConversionWarning(severity=lossy|dropped)`
+rather than silent drops — Cursor (no tool allowlist), Codex (only
+`sandbox_mode`), and Copilot's required `description` are the main offenders.
+
+Codex is the asymmetric case: its emitter returns a `CodexEmission`
+containing both the agent `.toml` body and a `[agents.<name>]` fragment for
+`~/.codex/config.toml`.  Orchestration features (`/fleet`, `/multitask`,
+Gemini's A2A `kind: remote`, Codex `max_depth`) are out of scope and
+documented as not translatable.
+
+CLI: `crossby agents convert --from <tool> --to <tool> <input>`.
+
+### Request Flow
+
+```
+CLI command
+  → service (e.g. run_sync)
+    → AI tool adapter (AbstractAITool — auto-registered via __init_subclass__)
+      → sync writer (AbstractSyncWriter, keyed by (tool_id, concern) in SyncRegistry)
+```
+
+### Key Concepts
+
+- **`AIToolID`** (`models/ai.py`) — a `StrEnum`. Works as both an enum member and a string key.
+- **`AbstractAITool`** (`ai_tools/base.py`) — every adapter subclasses this. Setting the `TOOL_ID` class variable auto-registers the adapter via `__init_subclass__` — no other file needs to change.
+- **`SyncRegistry`** (`sync/base.py`) — maps `(tool_id, concern)` → writer instance. Populated in `sync/__init__.py`; `run_sync()` orchestrates matching writers and collects `SyncResult`s.
+- **`SyncConcern`** — enumeration of what a writer handles: `RULES`, `AGENTS`, `SKILLS`, `PERMISSIONS`, `HOOKS`, `MCP`, `PLUGINS`. `PLUGINS` is detect-only — `run_sync()` injects findings via `sync/plugins.py` after the regular writer pass.
+- **Canonical agent IR** lives in `subagents/` (PR #46): `SubagentIR` plus one parser and one emitter per tool. `sync.agents._sync_translate` / `CodexAgentsWriter` delegate to `subagents.api.convert` for cross-tool translation; `ConversionWarning`s with `severity=lossy|dropped` are turned into `<!-- crossby:manual-fix -->` blocks by `_ir_body_with_manual_fix` before emit so the lossy edge surfaces inside the artifact, not just on the terminal.
+- **Canonical skill model** (`sync/agent_models.py`) — `SkillDefinition` is a tool-neutral dataclass plus `parse_markdown_skill` / `render_markdown_skill` / `translate_skill_for_target`. Skills use the same on-disk SKILL.md shape across every tool today, so the canonical layer exists only to attach manual-fix notes for fields the target tool doesn't honour (Claude `allowed-tools` on non-Claude targets).
+- **Manual-fix block** (`sync/manual_fix.py`) — when a writer can't faithfully translate a source field, it embeds a stable `<!-- crossby:manual-fix:start --> ... <!-- crossby:manual-fix:end -->` block in the rendered file. The block survives markdown rendering, sits inside TOML multi-line strings without escaping, and is replaced 1:1 on re-runs (`strip_manual_fix_blocks` + `append_manual_fix_block` + `find_manual_fix_blocks`).
+- **Cross-provider mappings** (`sync/translation.py`) — Claude↔Codex family table for `model`, family-aware `effort` bias, and `permissionMode` ↔ `sandbox_mode`. Used by both the agents writer and `crossby launch`'s `build_launch_command` for cross-provider model translation.
+- **Pre-write inspection** (`sync/plan.py`, `sync/validate.py`) — `--plan` summarizes a dry-run by concern + manual-fix count; `--doctor` adds validation findings and a coarse `high`/`medium`/`low` readiness rating; `--validate-target` re-parses every synced file (TOML / JSON parseability, agent required fields, skill frontmatter, MCP `command` on PATH, instruction file size).
+- **Persistent reports** (`sync/report.py`) — every real (non-dry-run) sync writes `.crossby/sync-report.md` with a portable `| Status | Item | Notes |` table. Statuses: `Added`, `Check before using`, `Not Added` — driven by `(action, file_path)` rather than message-substring matching.
+- **`.crossby.yml`** is loaded by `config/loader.py` into Pydantic v2 models. **Sync does not depend on it** — it reads each tool's native config directly from standard paths. The config is only consulted by `crossby launch` for defaults.
+- **Symlinks are always relative** (`os.path.relpath`, `config/linker.py`) so they survive repo moves.
+- **Sync is idempotent** — re-running on already-linked files is a no-op. Translate writers hash-compare rendered output before deciding `created` / `updated` / `skipped`.
+
+### Headless vs. interactive launches
+
+`build_launch_command` takes two distinct prompt-related parameters:
+- `prompt` — used for batch/headless invocations (one-shot runs, CI), gated by the tool's `headless_flag`.
+- `initial_message` — used for interactive sessions, placed as the first positional arg before any flags.
+
+Keep these separate when adding launch logic.
+
+## Adding a New AI Tool
+
+The adapter pattern is designed so adding a tool is a single-file change.
+
+1. Add the tool ID to `AIToolID` in `src/crossby/models/ai.py`.
+2. Create `src/crossby/ai_tools/<tool>.py` subclassing `AbstractAITool`:
+   - Set `TOOL_ID = AIToolID.<TOOL>` (this auto-registers the adapter).
+   - Implement `capabilities()` returning an `AIToolCapabilities` with at minimum `binary`, `display_name`, `model_flag`, `supports_*` booleans.
+   - Override the optional hooks that apply — e.g. `yolo_args()`, `effort_args()`, `trusted_dirs_args()`, `normalize_model_format()`, `resolve_effort_model()`, `initial_message_args()`.
+3. If the tool should participate in `crossby sync`, add writers under `src/crossby/sync/<concern>.py` for each concern it supports (see below) and register them in `sync/__init__.py`.
+4. If the tool should be a handoff **source**, override `locate_sessions()` and `read_session()` in the adapter.
+5. Add static model entries to `src/crossby/data/` if the tool has a known model catalog.
+6. Add tests under `tests/` — unit tests for the adapter, and integration tests for any sync writers.
+
+Adapters are imported via `src/crossby/ai_tools/__init__.py`. Make sure to add your import there so `__init_subclass__` runs.
+
+## Adding a New Sync Writer
+
+Sync writers live in `src/crossby/sync/<concern>.py` and subclass `AbstractSyncWriter` (see `sync/base.py`). Each writer:
+
+1. Sets `tool_id: AIToolID` and `concern: SyncConcern`.
+2. Implements `sync(data, project_root, *, dry_run, force) -> SyncResult`.
+3. Must be idempotent — re-running on unchanged state should return `action="skipped"` (with `file_path` set when the artifact is already in place; `file_path=None` means "nothing was synced for this concern", which the report renderer maps to `Not Added`).
+4. Must respect `dry_run` — compute the intended change but make no filesystem writes.
+5. On write conflicts, honor `force` (backup + overwrite) vs. raising.
+
+Register the instance in `src/crossby/sync/__init__.py` alongside the other writers. `SyncRegistry` enforces uniqueness by `(tool_id, concern)`.
+
+### Symlink, copy, or translate
+
+Writers that own file-tree concerns (rules, agents, skills) support up to three strategies via `SyncData.<concern>_strategy`:
+
+- **`symlink`** (default): create relative symlinks. Cheapest; edits propagate everywhere; only works when source and target use the same on-disk schema.
+- **`copy`**: physical copy with optional per-file rewrite (e.g. translate tool names `Bash`→`Shell` for Cursor). Used when the user wants a real file to commit, or when a marker on the source content would otherwise leak across schemas.
+- **`translate`**: agent writers delegate to `subagents.api.convert(from_tool, to_tool, content)` (canonical `SubagentIR` + per-tool parsers/emitters). Skill writers parse via `SkillDefinition` and call `translate_skill_for_target`. Both attach manual-fix notes for fields the target doesn't honour, render back to the target's on-disk shape, hash-compare for idempotency, and remove stale outputs whose source disappeared.
+
+When you add a new writer that handles one of these concerns, decide which strategies it supports, plumb each through `_sync_symlink` / `_sync_copy` / `_sync_translate` (see `agents.py` / `skills.py` for the existing pattern), and add tests for each strategy plus the no-op idempotent case.
+
+### Adding a manual-fix path
+
+If your writer translates a field that the target tool may not enforce or understand:
+
+1. Build a `ManualFixNote` with a short `category` (e.g. `permissionMode`, `allowed-tools`) and a user-facing `message`.
+2. Attach the note via `definition.with_notes([note])` on the canonical model.
+3. The renderer (`render_markdown_skill`, `render_markdown_agent`, `render_toml_agent`) appends a `<!-- crossby:manual-fix --> … <!-- /crossby:manual-fix -->` block at the bottom — no extra plumbing needed.
+
+Keep notes short and literal. Avoid Crossby-internal terminology in the message; users editing the file shouldn't need to know about `SubagentIR` or `SkillDefinition` to act on the note.
+
+## Tool Reference
+
+Crossby translates its unified CLI flags into each tool's native syntax. A dash (—) means the tool does not support that feature; crossby raises an error if you pass an explicit flag that the target tool doesn't support (e.g. `--yolo` with OpenCode).
+
+### Launch Flags
+
+| Crossby Flag  | Claude                             | Copilot           | Gemini                       | Codex                                      | OpenCode          | Cursor                     | VS Code | Antigravity |
+| ------------- | ---------------------------------- | ----------------- | ---------------------------- | ------------------------------------------ | ----------------- | -------------------------- | ------- | ----------- |
+| Binary        | `claude`                           | `copilot`         | `gemini`                     | `codex`                                    | `opencode`        | `agent`                    | `code`  | `antigravity` |
+| `--model`     | `--model`                          | `--model`         | `--model`                    | `--model`                                  | `--model`         | `--model`                  | —       | —           |
+| `--yolo`      | `--dangerously-skip-permissions`   | `--yolo`          | `--yolo`                     | `--yolo`                                   | —                 | `--force`                  | —       | —           |
+| `--plan`      | `--permission-mode plan`           | `--plan`          | `--approval-mode plan`       | —                                          | —                 | `--mode plan`              | —       | —           |
+| `--effort`    | `--effort <level>`                 | —                 | —                            | `-c model_reasoning_effort="…"`            | `--variant <level>` | model suffix (`-thinking`) | —       | —           |
+| `--prompt`    | positional                         | `-i <prompt>`     | positional                   | positional                                 | `--prompt <prompt>` | positional                | —       | —           |
+| `--transcript`| `script` wrapper                   | `script` wrapper  | `script` wrapper             | `script` wrapper                           | `script` wrapper  | `script` wrapper           | —       | —           |
+| `--resume`    | `--resume <id>`                    | `--resume=<id>`   | `--resume <id>`              | `codex resume <id>` (subcommand)           | `-s <id>`         | —                          | —       | —           |
+| `--trusted-dir` | `--add-dir`                      | `--add-dir`       | `--include-directories`      | `--sandbox workspace-write --add-dir`      | —                 | —                          | —       | —           |
+
+### Effort Level Mapping
+
+| Crossby Level | Claude   | Codex   | OpenCode | Cursor              |
+| ------------- | -------- | ------- | -------- | ------------------- |
+| `low`         | `low`    | `low`   | `low`    | —                   |
+| `medium`      | `medium` | `medium`| `medium` | —                   |
+| `high`        | `high`   | `high`  | `high`   | `<model>-thinking`  |
+| `max`         | `max`    | `xhigh` | `high`   | `<model>-thinking`  |
+
+### Permission & Allowlist Configuration
+
+Crossby stores canonical command patterns (e.g. `myapp:*`) and writes them into each tool's native config format.
+
+| Feature            | Claude                      | Copilot                        | Gemini                 | Cursor                        |
+| ------------------ | --------------------------- | ------------------------------ | ---------------------- | ----------------------------- |
+| Config file        | `.claude/settings.json`     | `.github/hooks/hooks.json`     | `.gemini/settings.json`| `.cursor/cli.json`            |
+| Allowlist format   | `Bash(cmd:args)`            | `shell(cmd:args)`              | `shell(cmd:args)`      | `Shell(cmd:args)`             |
+| Launch flag        | `--allowedTools`            | `--allow-tool`                 | `--allowed-tools`      | — (config-file only)          |
+| Hook config        | `hooks.PreToolUse`          | `hooks.preToolUse`             | `hooks.BeforeTool`     | `preToolUse` in `hooks.json`  |
+| Hook guard matcher | `Edit\|Write\|NotebookEdit` | `Write\|Delete`                | file-write tools       | `Write\|Delete`               |
+
+### Session Preservation & Resume
+
+| Feature                 | Claude                  | Copilot             | Gemini          | Codex              | OpenCode   | Cursor                   |
+| ----------------------- | ----------------------- | ------------------- | --------------- | ------------------ | ---------- | ------------------------ |
+| Resume command          | `claude --resume <id>`  | `copilot --resume=<id>` | `gemini --resume <id>` | `codex resume <id>` | `opencode -s <id>` | — |
+| Session data path       | `~/.claude/projects/`   | —                   | —               | —                  | —          | `~/.cursor/projects/`    |
+| Session data preserved  | Yes (worktree → main)   | —                   | —               | —                  | —          | Yes (worktree → main)    |
+
+Session IDs are extracted automatically from transcripts when `--transcript` is used.
+
+### Transcript Parsing (`crossby stats`)
+
+| Feature                  | Claude | Copilot | Gemini | Codex |
+| ------------------------ | ------ | ------- | ------ | ----- |
+| Total tokens             | Yes    | Yes     | Yes    | Yes   |
+| Input / output breakdown | Yes    | Yes     | Yes    | Yes   |
+| Cached tokens            | Yes    | Yes     | Yes    | Yes   |
+| Per-model breakdown      | —      | Yes     | Yes    | —     |
+| Premium requests         | —      | Yes     | —      | —     |
+| Session ID extraction    | Yes    | Yes     | Yes    | Yes   |
+
+### Handoff Sources & Targets
+
+| Tool                                  | Source (read)                                        | Target (launch) |
+| ------------------------------------- | ---------------------------------------------------- | --------------- |
+| Claude                                | ✓ (`~/.claude/projects/<encoded>/<id>.jsonl`)        | ✓               |
+| Cursor                                | ✓ (`~/.cursor/projects/<encoded>/chat.json`)         | ✓               |
+| Codex                                 | ✓ (`~/.codex/sessions/YYYY/MM/DD/rollout-*.jsonl`)   | ✓               |
+| Copilot                               | ✓ (`~/.copilot/session-state/<id>/events.jsonl`)     | ✓               |
+| Gemini, OpenCode, Antigravity, VS Code| —                                                    | ✓               |
+
+### Library API (not exposed via CLI)
+
+Available on each adapter for programmatic use:
+
+| Feature           | Claude                                   | Copilot             | Gemini                   | Codex        | OpenCode           | Cursor          |
+| ----------------- | ---------------------------------------- | ------------------- | ------------------------ | ------------ | ------------------ | --------------- |
+| Trusted dirs      | `--add-dir`                              | `--add-dir`         | `--include-directories`  | `--add-dir`  | —                  | —               |
+| Structured output | `--output-format json --json-schema …`   | —                   | `--output-format json`   | —            | —                  | —               |
+| Model format      | dashed (`claude-haiku-4-5`)              | dotted (`claude-haiku-4.5`) | as-is            | as-is        | `provider/model`   | as-is           |
+
+## Commit Conventions
+
+This project uses [Conventional Commits](https://www.conventionalcommits.org/):
+
+```
+<type>(<scope>): <short summary>
+
+<optional body>
+
+<optional footer>
+```
+
+Common types: `feat`, `fix`, `refactor`, `test`, `docs`, `chore`.
+
+Examples:
+```
+feat(sync): add allowlist conversion for Gemini
+fix(cli): handle missing .crossby.yml gracefully
+docs: update compatibility table for Codex effort levels
+```
+
+Breaking changes: append `!` after the type (e.g. `feat!:`) and add a `BREAKING CHANGE:` footer.
+
+## Release Process
+
+1. Ensure `./scripts/check-all.sh` passes on `main`.
+2. Update the version in `pyproject.toml`.
+3. Commit: `chore: release vX.Y.Z`.
+4. Tag: `git tag vX.Y.Z`.
+5. Push both: `git push origin main && git push origin vX.Y.Z`.

crossby-0.2.4/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2024 Ivan Viragine
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.