@iowarp/clio-coder 0.4.6 → 0.4.7

package/library/plugins/materio/assets/scripts/project_plugin.py

@@ -0,0 +1,165 @@
+#!/usr/bin/env python3
+"""Generate a peer host package without claiming equivalent fleet enforcement."""
+import argparse
+import json
+from pathlib import Path
+import shutil
+import sys
+
+
+ACTION_ROLES = {
+    "identify-research": "research-explorer", "literature-review": "literature-reviewer",
+    "define-virtual-lab": "lab-definer", "define-research-tasks": "workflow-planner",
+    "execute-task": "task-executor",
+}
+
+
+def render_peer_guides(target, output, name):
+    """Bind logical action/skill requests to capabilities actually exported."""
+    actions = sorted(path.stem for path in (output / "assets/actions").glob("*.md"))
+    invocations = {}
+    for action in actions:
+        skill = name + "-" + ACTION_ROLES.get(action, "research-explorer")
+        invocations[action] = ({"kind": "skill", "name": skill, "request": action}
+                               if target == "codex" else {"kind": "command", "name": f"/{name}:{action}"})
+    paths = list((output / "assets/actions").glob("*.md")) + list((output / "skills").glob("*/SKILL.md"))
+    for path in paths:
+        text = path.read_text()
+        if target in {"claude", "gemini"}:
+            for action in actions:
+                skill = name + "-" + ACTION_ROLES.get(action, "research-explorer")
+                text = text.replace(f"Run {action} using the {skill} skill", f"/{name}:{action}")
+            text = text.replace("this standard package registers no slash commands",
+                                "this native package registers the command routes shown below")
+            text = text.replace("actual skill invocation vocabulary", "native command vocabulary")
+        if path.parent.name == "actions":
+            role = ACTION_ROLES.get(path.stem)
+            if target == "claude" and role:
+                binding = (f"\n## Native role binding\n\nThe interactive orchestrator dispatches the registered agent `{name}:{role}`\n"
+                           f"for the `{name}-{role}` scientific skill. After a checkpoint, dispatch that\n"
+                           "same registered agent again with the complete transcript, prior outputs and\n"
+                           "actual researcher answer. Skill names identify instructions, not agent IDs.\n"
+                           "A worker returns the checkpoint and does not dispatch or interview.\n")
+                if role == "task-executor":
+                    binding += (f"For the separate read-only artifact review, dispatch the registered agent `{name}:task-verifier`.\n"
+                                "Give it no write grant and preserve any host grounding limitations.\n")
+            elif target in {"codex", "gemini"}:
+                binding = ("\n## Host role execution\n\nThis export registers no named agents. Apply the role's installed skill in the\n"
+                           "main conversation, or give that skill and complete context to a generic worker\n"
+                           "only when the host supports delegation. The interactive assistant owns every\n"
+                           "researcher question and resumes the role with the actual answer.\n")
+            else:
+                binding = ""
+            text += binding
+        path.write_text(text)
+    return invocations
+
+
+def project(target, output, source=None):
+    source = Path(source or Path(__file__).resolve().parents[2]).resolve(strict=True)
+    output = Path(output).absolute()
+    if output.exists() or output.is_symlink():
+        raise ValueError("projection output must not already exist")
+    output = output.resolve()
+    if source == output or source in output.parents:
+        raise ValueError("projection output must be outside the canonical bundle")
+    manifest = json.loads((source / "plugin.json").read_text())
+    output.mkdir(parents=True)
+    try:
+        for directory in ("skills", "assets"):
+            for path in (source / directory).rglob("*"):
+                if path.is_symlink():
+                    raise ValueError(f"projection refuses symbolic link: {path}")
+            shutil.copytree(source / directory, output / directory,
+                            ignore=shutil.ignore_patterns("__pycache__", "*.pyc"))
+        metadata = {key: manifest[key] for key in ("name", "version", "description", "author", "repository", "keywords") if key in manifest}
+        report = {"target": target, "skills": 6, "commands": 0, "agents": 0,
+                  "fleet": "instructions only; no native write-scope enforcement claimed",
+                  "scripts": "local Python helpers; no automatic execution",
+                  "validation": "generated structure; run target-native validation before publication"}
+        report["invocations"] = render_peer_guides(target, output, metadata["name"])
+        report["agent_bindings"] = {}
+        if target == "codex":
+            # Codex recognizes the current portable standard. Do not fabricate a
+            # separate proprietary manifest or pretend Clio recipes are native.
+            portable = {"$schema": manifest["$schema"], **metadata}
+            (output / "plugin.json").write_text(json.dumps(portable, indent=2) + "\n")
+        elif target == "claude":
+            directory = output / ".claude-plugin"
+            directory.mkdir()
+            (directory / "plugin.json").write_text(json.dumps(metadata, indent=2) + "\n")
+            commands = output / "commands"
+            agents = output / "agents"
+            commands.mkdir()
+            agents.mkdir()
+            roles = ACTION_ROLES
+            for guide in sorted((output / "assets/actions").glob("*.md")):
+                assignment = (f"Delegate this scientific role to the registered agent `{metadata['name']}:{roles[guide.stem]}`.\n"
+                              if guide.stem in roles else "This action is owned by the interactive orchestrator.\n")
+                text = (f"---\ndescription: Materials characterization {guide.stem} workflow.\n---\n\n"
+                        f"Read ${{CLAUDE_PLUGIN_ROOT}}/assets/actions/{guide.name} and follow that workflow.\n"
+                        + assignment +
+                        "Use AskUserQuestion for the interactive interview. If delegating, use the registered\n"
+                        "materio role agent and include full context; workers return\n"
+                        "checkpoint questions to this orchestrator. Use Bash only for approved local helper\n"
+                        "calls; bind PACKAGE_ROOT to ${CLAUDE_PLUGIN_ROOT} in each call.\n"
+                        "Arguments: $ARGUMENTS\n")
+                (commands / guide.name).write_text(text)
+                report["commands"] += 1
+            for skill in sorted((output / "skills").glob("*/SKILL.md")):
+                role = skill.parent.name.removeprefix("materio-")
+                tools = "Read, Glob, Grep" if role == "task-verifier" else "Read, Glob, Grep, Write, Edit, Bash, WebFetch"
+                text = (f"---\nname: {role}\ndescription: Materials characterization {role} role.\n"
+                        f"tools: {tools}\n---\n\n"
+                        f"Read ${{CLAUDE_PLUGIN_ROOT}}/skills/{skill.parent.name}/SKILL.md.\n"
+                        "Apply its scientific contract within the caller's approved output paths.\n"
+                        "Return structured complete, blocked, or needs_input with checkpoint kind\n"
+                        "decision, human-action, or human-verify. Do not interview the researcher.\n"
+                        "The caller owns state reconciliation, optional recording and paper gates.\n")
+                (agents / (role + ".md")).write_text(text)
+                report["agents"] += 1
+                report["agent_bindings"][role] = metadata["name"] + ":" + role
+        elif target == "gemini":
+            (output / "gemini-extension.json").write_text(json.dumps({
+                "name": metadata["name"], "version": metadata["version"],
+                "description": metadata["description"], "contextFileName": "GEMINI.md"
+            }, indent=2) + "\n")
+            (output / "GEMINI.md").write_text(
+                "# Materials characterization\n\nUse the installed materio skills and their relative action-guide links.\n"
+                "The interactive assistant owns researcher questions and state changes. Roles may\n"
+                "run inline when delegation is unavailable. No native fleet boundary is claimed.\n")
+            directory = output / "commands/materio"
+            directory.mkdir(parents=True)
+            for guide in sorted((output / "assets/actions").glob("*.md")):
+                text = (f"Use the installed materio skills to perform {guide.stem}. "
+                        f"Read their relative link to assets/actions/{guide.name} and follow its contract. "
+                        "Use the host conversation tool for researcher interviews. Delegate only with "
+                        "available capabilities; otherwise perform the role inline. Arguments: {{args}}")
+                (directory / (guide.stem + ".toml")).write_text(
+                    "description = " + json.dumps(f"Materials characterization {guide.stem}") + "\n"
+                    "prompt = " + json.dumps(text) + "\n")
+                report["commands"] += 1
+        else:
+            raise ValueError(f"unsupported peer target: {target}")
+        (output / "capability-report.json").write_text(json.dumps(report, indent=2) + "\n")
+        return report
+    except Exception:
+        shutil.rmtree(output)
+        raise
+
+
+def main():
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument("--target", choices=["codex", "claude", "gemini"], required=True)
+    parser.add_argument("--output", required=True)
+    args = parser.parse_args()
+    try:
+        print(json.dumps(project(args.target, args.output), indent=2))
+        return 0
+    except (OSError, ValueError) as error:
+        print(str(error), file=sys.stderr)
+        return 2
+
+
+if __name__ == "__main__":
+    sys.exit(main())

package/library/plugins/materio/assets/scripts/research_state.py

@@ -0,0 +1,400 @@
+#!/usr/bin/env python3
+"""Deterministic local research state operations. Python 3.10+, stdlib only.
+
+This helper never initializes git, tags a checkpoint, or writes paper state.
+It does not run experiments or generated analysis programs.
+"""
+import argparse
+import hashlib
+import io
+import json
+import os
+from pathlib import Path
+import re
+import shutil
+import subprocess
+import sys
+import tarfile
+import tempfile
+from datetime import datetime, timezone
+from uuid import uuid4
+
+DEFAULTS = {"web_search": False, "auto_checkpoint": True, "commit_research": False}
+PRESERVED = {"data", "checkpoints"}
+ARCHIVE_MANIFEST = ".snapshot-manifest.json"
+
+
+def unique_object(pairs):
+    result = {}
+    for key, value in pairs:
+        if key in result:
+            raise ValueError(f"duplicate JSON key: {key}")
+        result[key] = value
+    return result
+
+
+def research_root(project):
+    project = Path(project).resolve(strict=True)
+    root = project / ".research"
+    if root.is_symlink():
+        raise ValueError(".research must be a real directory, not a symlink")
+    return root
+
+
+def config(root):
+    path = root / "config.json"
+    result = dict(DEFAULTS)
+    if path.is_symlink():
+        raise ValueError("config.json must not be a symlink")
+    if path.exists():
+        value = json.loads(path.read_text(), object_pairs_hook=unique_object)
+        if not isinstance(value, dict):
+            raise ValueError("research config must be a JSON object")
+        for key in DEFAULTS:
+            if key in value and type(value[key]) is not bool:
+                raise ValueError(f"{key} must be a JSON boolean")
+        result.update(value)
+    return result
+
+
+def contained(root, relative, allow_missing=False):
+    relative = Path(relative)
+    if relative.is_absolute() or ".." in relative.parts or not relative.parts:
+        raise ValueError(f"invalid relative research path: {relative}")
+    path = root
+    for part in relative.parts:
+        path = path / part
+        if path.is_symlink():
+            raise ValueError(f"symlink is not permitted: {path}")
+    if not allow_missing and not path.exists():
+        raise ValueError(f"missing research path: {path}")
+    return path
+
+
+def task_number(value):
+    if not re.fullmatch(r"[0-9]+", str(value)) or int(value) < 1:
+        raise ValueError("task must be a positive decimal integer")
+    return f"{int(value):02d}"
+
+
+def workflow(root):
+    path = root / "WORKFLOW.md"
+    if not path.exists():
+        return [], []
+    text = contained(root, "WORKFLOW.md").read_text()
+    active_text, _, archive_text = text.partition("## Archived Tasks")
+    def parse(text):
+        matches = list(re.finditer(r"^### Task ([0-9]+):[^\n]*", text, re.M))
+        return [(task_number(m.group(1)), text[m.start():matches[i + 1].start() if i + 1 < len(matches) else len(text)]) for i, m in enumerate(matches)]
+    active, archived = parse(active_text), parse(archive_text)
+    ids = [number for number, _ in active + archived]
+    if len(set(ids)) != len(ids):
+        raise ValueError("duplicate task identity in WORKFLOW.md")
+    return active, archived
+
+
+def next_task(root):
+    active, archived = workflow(root)
+    ids = [int(number) for number, _ in active + archived]
+    tasks = root / "tasks"
+    if tasks.exists():
+        contained(root, "tasks")
+        for path in tasks.iterdir():
+            m = re.fullmatch(r"task-([0-9]+)", path.name)
+            if m:
+                ids.append(int(m.group(1)))
+    return task_number(max(ids, default=0) + 1)
+
+
+def validate_workflow(root):
+    active, archived = workflow(root)
+    ids = {number for number, _ in active}
+    archived_ids = {number for number, _ in archived}
+    graph = {}
+    for number, block in active:
+        match = re.search(r"^[-*] \*\*Dependencies\*\*:\s*(.*)$", block, re.M)
+        refs = [task_number(value) for value in re.findall(r"Task\s+([0-9]+)", match.group(1) if match else "")]
+        for ref in refs:
+            if ref not in ids:
+                kind = "archived" if ref in archived_ids else "missing"
+                raise ValueError(f"Task {number} depends on {kind} Task {ref}")
+        graph[number] = refs
+    visiting, visited = set(), set()
+    def visit(number):
+        if number in visiting:
+            raise ValueError(f"dependency cycle at Task {number}")
+        if number in visited:
+            return
+        visiting.add(number)
+        for ref in graph[number]:
+            visit(ref)
+        visiting.remove(number)
+        visited.add(number)
+    for number in graph:
+        visit(number)
+    return {"active": list(graph), "archived": sorted(archived_ids), "dependencies": graph}
+
+
+def next_ready_task(root):
+    """Choose a pending/in-progress task by workflow order and completed inputs."""
+    graph = validate_workflow(root)["dependencies"]
+    active, _ = workflow(root)
+    statuses = {}
+    for number, block in active:
+        match = re.search(r"^[-*] \*\*Status\*\*:\s*(.*)$", block, re.M)
+        status = re.sub(r"^[^a-zA-Z]+", "", match.group(1) if match else "").lower()
+        statuses[number] = status.split()[0] if status else "unspecified"
+    blocked = []
+    for number, _ in active:
+        if statuses[number] not in {"pending", "in-progress"}:
+            continue
+        unsettled = [dependency for dependency in graph[number] if statuses[dependency] != "complete"]
+        if not unsettled:
+            return {"id": number, "status": statuses[number], "blocked": blocked}
+        blocked.append({"id": number, "dependencies": unsettled})
+    return {"id": None, "reason": "no dependency-ready pending task", "blocked": blocked}
+
+
+def snapshot_files(root):
+    files = {}
+    for base, directories, names in os.walk(root, followlinks=False):
+        base = Path(base)
+        if base == root:
+            directories[:] = [name for name in directories if name not in PRESERVED]
+        for name in directories:
+            if (base / name).is_symlink():
+                raise ValueError(f"snapshot refuses symlink: {base / name}")
+        for name in names:
+            path = base / name
+            relative = path.relative_to(root).as_posix()
+            if base == root and name in PRESERVED:
+                continue
+            if path == root / ARCHIVE_MANIFEST:
+                raise ValueError(f"reserved snapshot filename: {ARCHIVE_MANIFEST}")
+            if path.is_symlink() or not path.is_file():
+                raise ValueError(f"snapshot needs regular files: {path}")
+            files[relative] = path.read_bytes()
+    return files
+
+
+def checkpoint_save(root, label):
+    if not re.fullmatch(r"[a-zA-Z0-9][a-zA-Z0-9._-]{0,63}", label):
+        raise ValueError("checkpoint label must be 1–64 path-safe characters")
+    if not root.is_dir():
+        raise ValueError("research state is not initialized")
+    directory = contained(root, "checkpoints", allow_missing=True)
+    directory.mkdir(exist_ok=True)
+    files = snapshot_files(root)
+    directories = []
+    for base, names, _ in os.walk(root, followlinks=False):
+        if Path(base) == root:
+            names[:] = [name for name in names if name not in PRESERVED]
+        directories.extend((Path(base) / name).relative_to(root).as_posix() for name in names)
+    modes = {name: (root / name).stat().st_mode & 0o777 for name in files}
+    manifest = {"version": 1, "files": {name: hashlib.sha256(data).hexdigest() for name, data in files.items()},
+                "directories": sorted(directories), "modes": modes}
+    name = datetime.now(timezone.utc).strftime("%Y%m%dT%H%M%S.%fZ") + "-" + label + "-" + uuid4().hex[:8] + ".tgz"
+    destination = directory / name
+    created = False
+    try:
+        with destination.open("xb") as output:
+            created = True
+            with tarfile.open(fileobj=output, mode="w:gz") as archive:
+                for relative, data in sorted({**files, ARCHIVE_MANIFEST: json.dumps(manifest).encode()}.items()):
+                    member = tarfile.TarInfo(relative)
+                    member.size, member.mode = len(data), 0o600
+                    archive.addfile(member, io.BytesIO(data))
+    except Exception:
+        if created:
+            destination.unlink(missing_ok=True)
+        raise
+    return {"checkpoint": name, "files": len(files), "excluded": sorted(PRESERVED)}
+
+
+def checkpoint_read(root, name):
+    if not re.fullmatch(r"[a-zA-Z0-9][a-zA-Z0-9._-]*\.tgz", name):
+        raise ValueError("checkpoint must be a filename listed by checkpoint list")
+    path = contained(root, "checkpoints/" + name)
+    files = {}
+    with tarfile.open(path, "r:gz") as archive:
+        for member in archive:
+            relative = Path(member.name)
+            if member.name in files or relative.is_absolute() or ".." in relative.parts or not relative.parts:
+                raise ValueError(f"invalid or duplicate archive member: {member.name}")
+            if relative.parts[0] in PRESERVED or not member.isfile():
+                raise ValueError(f"archive member is not permitted: {member.name}")
+            if relative.as_posix() != member.name:
+                raise ValueError(f"noncanonical archive member: {member.name}")
+            files[member.name] = archive.extractfile(member).read()
+    if ARCHIVE_MANIFEST not in files:
+        raise ValueError("checkpoint has no integrity manifest")
+    manifest = json.loads(files.pop(ARCHIVE_MANIFEST), object_pairs_hook=unique_object)
+    observed = {key: hashlib.sha256(value).hexdigest() for key, value in files.items()}
+    if not isinstance(manifest, dict) or manifest.get("version") != 1 or manifest.get("files") != observed:
+        raise ValueError("checkpoint content integrity mismatch")
+    directories = manifest.get("directories", [])
+    modes = manifest.get("modes", {})
+    if not isinstance(directories, list) or not isinstance(modes, dict):
+        raise ValueError("invalid snapshot directory or mode inventory")
+    for directory in directories:
+        if not isinstance(directory, str):
+            raise ValueError("invalid snapshot directory")
+        relative = Path(directory)
+        if relative.is_absolute() or ".." in relative.parts or not relative.parts or relative.parts[0] in PRESERVED or relative.as_posix() != directory:
+            raise ValueError(f"invalid snapshot directory: {directory}")
+    if any(name not in files or type(mode) is not int or mode < 0 or mode > 0o777 for name, mode in modes.items()):
+        raise ValueError("invalid snapshot mode")
+    return files, directories, modes
+
+
+def checkpoint_restore(root, name, confirmed):
+    if not confirmed:
+        raise ValueError("restore requires --confirmed after researcher confirmation")
+    archive_name = name
+    files, directories, modes = checkpoint_read(root, name)
+    # Validate and stage before touching current state. Data and checkpoints are
+    # moved, never traversed or copied, and rollback restores the original root.
+    stage = Path(tempfile.mkdtemp(prefix=".research-restore-", dir=root.parent))
+    backup = Path(tempfile.mkdtemp(prefix=".research-previous-", dir=root.parent))
+    backup.rmdir()
+    moved = []
+    swapped = False
+    try:
+        for directory in directories:
+            (stage / directory).mkdir(parents=True, exist_ok=True)
+        for relative, data in files.items():
+            path = stage / relative
+            path.parent.mkdir(parents=True, exist_ok=True)
+            path.write_bytes(data)
+            path.chmod(modes.get(relative, 0o600))
+        root.rename(backup)
+        swapped = True
+        for name in sorted(PRESERVED):
+            if (backup / name).exists() or (backup / name).is_symlink():
+                (backup / name).rename(stage / name)
+                moved.append(name)
+        stage.rename(root)
+    except Exception:
+        if swapped:
+            for name in reversed(moved):
+                (stage / name).rename(backup / name)
+            backup.rename(root)
+        raise
+    finally:
+        if stage.exists():
+            shutil.rmtree(stage)
+    shutil.rmtree(backup)
+    return {"restored": archive_name, "files": len(files), "preserved": sorted(PRESERVED)}
+
+
+def record(root, message, paths):
+    if not config(root)["commit_research"]:
+        return {"recorded": False, "reason": "commit_research is false"}
+    project = root.parent
+    def git(*argv, check=True):
+        return subprocess.run(["git", "-C", str(project), *argv], text=True, capture_output=True, check=check)
+    git("rev-parse", "--show-toplevel")
+    if git("diff", "--cached", "--name-only").stdout.strip():
+        raise ValueError("optional record refused: unrelated or preexisting staged changes")
+    selected = []
+    for value in paths:
+        relative = Path(value)
+        if not relative.parts or relative.parts[0] != ".research":
+            raise ValueError("optional record accepts only explicit .research files")
+        path = contained(root, Path(*relative.parts[1:]))
+        if not path.is_file():
+            raise ValueError("record needs named files, not directory staging")
+        selected.append(relative.as_posix())
+    if not selected:
+        return {"recorded": False, "reason": "no changed files supplied"}
+    git("add", "--", *selected)
+    if git("diff", "--cached", "--quiet", check=False).returncode == 0:
+        return {"recorded": False, "reason": "no staged changes"}
+    result = git("commit", "-m", message)
+    return {"recorded": True, "files": selected, "detail": result.stdout.strip()}
+
+
+def register_copy(root, source):
+    source = Path(source).resolve(strict=True)
+    if not source.is_file():
+        raise ValueError("data registration needs an actual file")
+    directory = contained(root, "data", allow_missing=True)
+    directory.mkdir(exist_ok=True)
+    target = directory / source.name
+    data = source.read_bytes()
+    digest = hashlib.sha256(data).hexdigest()
+    if target.exists() or target.is_symlink():
+        if target.is_file() and not target.is_symlink() and target.read_bytes() == data:
+            return {"path": str(target), "source": str(source), "sha256": digest, "copied": False}
+        raise ValueError(f"data filename collision: {target.name}; choose a distinct name")
+    with target.open("xb") as output:
+        output.write(data)
+    return {"path": str(target), "source": str(source), "sha256": digest, "copied": True}
+
+
+def main(argv=None):
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument("--project", default=".")
+    sub = parser.add_subparsers(dest="action", required=True)
+    sub.add_parser("init")
+    cfg = sub.add_parser("config")
+    cfg.add_argument("--key", choices=sorted(DEFAULTS))
+    sub.add_parser("next-task-id")
+    sub.add_parser("next-task")
+    sub.add_parser("validate-workflow")
+    sub.add_parser("task-dirs")
+    checkpoint = sub.add_parser("checkpoint")
+    checkpoint.add_argument("operation", choices=["save", "restore", "list"])
+    checkpoint.add_argument("name", nargs="?", default="manual")
+    checkpoint.add_argument("--confirmed", action="store_true")
+    rec = sub.add_parser("record")
+    rec.add_argument("--message", required=True)
+    rec.add_argument("--files", nargs="+", required=True)
+    copy = sub.add_parser("copy-data")
+    copy.add_argument("source")
+    args = parser.parse_args(argv)
+    try:
+        root = research_root(args.project)
+        if args.action == "init":
+            config(root)
+            root.mkdir(exist_ok=True)
+            for name in ("tasks", "data", "checkpoints"):
+                contained(root, name, allow_missing=True).mkdir(exist_ok=True)
+            if not (root / "config.json").exists():
+                with (root / "config.json").open("x") as output:
+                    output.write(json.dumps(DEFAULTS, indent=2) + "\n")
+            result = {"initialized": True, "config": config(root)}
+        elif args.action == "config":
+            result = config(root)
+            if args.key:
+                result = result[args.key]
+        elif args.action == "next-task-id":
+            result = {"id": next_task(root)}
+        elif args.action == "next-task":
+            result = next_ready_task(root)
+        elif args.action in ("validate-workflow", "task-dirs"):
+            result = validate_workflow(root)
+            if args.action == "task-dirs":
+                for number in result["active"]:
+                    contained(root, f"tasks/task-{number}", allow_missing=True).mkdir(parents=True, exist_ok=True)
+        elif args.action == "checkpoint":
+            if args.operation == "save":
+                result = checkpoint_save(root, args.name)
+            elif args.operation == "restore":
+                result = checkpoint_restore(root, args.name, args.confirmed)
+            else:
+                directory = contained(root, "checkpoints", allow_missing=True)
+                result = {"checkpoints": sorted(p.name for p in directory.glob("*.tgz") if p.is_file() and not p.is_symlink())}
+        elif args.action == "record":
+            result = record(root, args.message, args.files)
+        else:
+            result = register_copy(root, args.source)
+        print(json.dumps(result, indent=2))
+        return 0
+    except (ValueError, OSError, json.JSONDecodeError, tarfile.TarError, subprocess.CalledProcessError) as error:
+        print(json.dumps({"error": str(error)}), file=sys.stderr)
+        return 2
+
+
+if __name__ == "__main__":
+    sys.exit(main())