@yottameta/yotta-dev-mcp-plugin 0.1.1 → 0.2.1

package/skills/yotta-dev-mcp/scripts/dev_selftest.py

@@ -0,0 +1,544 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""Self-test and counterexample probes for yotta-dev-mcp."""
+
+import ast
+import hashlib
+import json
+import os
+import re
+import shutil
+import subprocess
+import sys
+import tempfile
+from pathlib import Path
+
+from dev_architecture import architecture_review
+from dev_common import (
+    EVIDENCE_LIMIT, LEAN_INSTALL_MARKERS, VERIFY_REQUIRED_INSTALLED_ASSETS,
+    VERIFY_REQUIRED_INSTALLED_FILES,
+    VERIFY_REQUIRED_SOURCE_FILES, VERIFY_WRITE_GATES, _frontmatter_version,
+    _read_text,
+)
+
+
+def _function_default(source_text, function_name, parameter_name):
+    try:
+        tree = ast.parse(source_text)
+    except SyntaxError:
+        return False, None
+    for node in tree.body:
+        if not isinstance(node, ast.FunctionDef) or node.name != function_name:
+            continue
+        positional = list(node.args.args)
+        defaults = list(node.args.defaults)
+        offset = len(positional) - len(defaults)
+        for index, argument in enumerate(positional):
+            if argument.arg != parameter_name or index < offset:
+                continue
+            try:
+                return True, ast.literal_eval(defaults[index - offset])
+            except (ValueError, SyntaxError):
+                return True, "<non-literal>"
+    return False, None
+
+def _string_constant(node):
+    return isinstance(node, ast.Constant) and isinstance(node.value, str)
+
+def _is_lean_install(root):
+    return any((root / marker).is_file() for marker in LEAN_INSTALL_MARKERS)
+
+def _protocol_tool_contracts(source_text):
+    tree = ast.parse(source_text)
+    function = None
+    for node in tree.body:
+        if isinstance(node, ast.FunctionDef) and node.name == "mcp_tools":
+            function = node
+            break
+    if function is None:
+        raise ValueError("mcp_tools() not found")
+    list_node = None
+    for node in ast.walk(function):
+        if isinstance(node, ast.Return) and isinstance(node.value, ast.List):
+            list_node = node.value
+            break
+    if list_node is None:
+        raise ValueError("mcp_tools() does not return a literal tool list")
+    contracts = []
+    for element in list_node.elts:
+        if not isinstance(element, ast.Dict):
+            raise ValueError("mcp_tools() contains a non-literal tool entry")
+        values = {}
+        for key, value in zip(element.keys, element.values):
+            if _string_constant(key):
+                values[key.value] = value
+        name_node = values.get("name")
+        if name_node is None:
+            raise ValueError("tool entry is missing name")
+        try:
+            name = ast.literal_eval(name_node)
+        except (ValueError, SyntaxError):
+            raise ValueError("tool name is not a literal string")
+        schema = values.get("inputSchema")
+        additional = False
+        if isinstance(schema, ast.Dict):
+            for key, value in zip(schema.keys, schema.values):
+                if _string_constant(key) and key.value == "additionalProperties":
+                    try:
+                        additional = ast.literal_eval(value) is False
+                    except (ValueError, SyntaxError):
+                        additional = False
+        contracts.append({"name": name, "additional_properties": additional})
+    return contracts
+
+def _dispatch_tool_names(source_text):
+    tree = ast.parse(source_text)
+    for node in tree.body:
+        if not isinstance(node, ast.FunctionDef) or node.name != "dispatch":
+            continue
+        for child in ast.walk(node):
+            if not isinstance(child, ast.Assign):
+                continue
+            if not any(isinstance(target, ast.Name) and target.id == "handlers"
+                       for target in child.targets):
+                continue
+            if not isinstance(child.value, ast.Dict):
+                continue
+            names = []
+            for key in child.value.keys:
+                if _string_constant(key):
+                    names.append(key.value)
+                else:
+                    try:
+                        names.append(ast.literal_eval(key))
+                    except (ValueError, SyntaxError):
+                        pass
+            return sorted(names)
+    raise ValueError("dispatch() handler map not found")
+
+def _self_test_counterexamples():
+    probes = []
+
+    def record(kind, name, expectation, observed, passed, detail):
+        probes.append({
+            "kind": kind,
+            "name": name,
+            "expectation": expectation,
+            "observed": observed,
+            "passed": bool(passed),
+            "detail": detail,
+        })
+
+    with tempfile.TemporaryDirectory(prefix="yotta-dev-mcp-selftest-") as tmp:
+        root = Path(tmp)
+
+        def write(rel, text):
+            target = root / rel
+            target.parent.mkdir(parents=True, exist_ok=True)
+            target.write_text(text, encoding="utf-8")
+
+        contract = {
+            "version": 1,
+            "layers": [
+                {"id": "core", "paths": ["core/**"]},
+                {"id": "ui", "paths": ["ui/**"]},
+            ],
+            "rules": [{
+                "id": "core-no-ui",
+                "type": "forbid-dependency",
+                "from": "core",
+                "to": "ui",
+                "severity": "high",
+                "claim": "core must not import ui",
+            }],
+        }
+        write(".yotta/architecture.json", json.dumps(contract))
+        write("core/leak.py", "from ui.view import VALUE\n")
+        write("ui/view.py", "VALUE = 1\n")
+
+        try:
+            observed = architecture_review(str(root))["status"]
+            detail = "seeded core -> ui dependency was reviewed"
+        except Exception as exc:  # noqa: BLE001
+            observed = "error"
+            detail = str(exc)
+        record("seeded-defect", "forbidden dependency", "FAIL", observed,
+               observed == "FAIL", detail)
+
+        contract["rules"] = []
+        write(".yotta/architecture.json", json.dumps(contract))
+        try:
+            observed = architecture_review(str(root))["status"]
+            detail = "same defect with the rule removed"
+        except Exception as exc:  # noqa: BLE001
+            observed = "error"
+            detail = str(exc)
+        record("mutation-control", "remove the rule", "not FAIL", observed,
+               observed != "FAIL", detail)
+
+        contract["version"] = 3
+        write(".yotta/architecture.json", json.dumps(contract))
+        try:
+            observed = architecture_review(str(root))["status"]
+            detail = "unsupported contract version"
+        except Exception as exc:  # noqa: BLE001
+            observed = "error"
+            detail = str(exc)
+        record("invalid-contract", "unsupported version", "FAIL or UNKNOWN",
+               observed, observed in ("FAIL", "UNKNOWN"), detail)
+
+        (root / ".yotta" / "architecture.json").unlink()
+        try:
+            observed = architecture_review(str(root))["status"]
+            detail = "missing contract"
+        except Exception as exc:  # noqa: BLE001
+            observed = "error"
+            detail = str(exc)
+        record("missing-contract", "no architecture contract", "UNKNOWN",
+               observed, observed == "UNKNOWN", detail)
+
+        write("secret.env", "TOKEN=abcdefghijklmnopqrstuvwxyz123456\n")
+        try:
+            from dev_engine import scan_secrets
+            observed = len(scan_secrets(str(root))["findings"])
+            passed = observed >= 1
+            detail = "seeded token was scanned"
+        except Exception as exc:  # noqa: BLE001
+            observed = "error"
+            passed = False
+            detail = str(exc)
+        record("secret-scan", "seeded credential", "at least one finding",
+               observed, passed, detail)
+
+        write("package.json", json.dumps({"name": "probe", "version": "1.0.0"}))
+        write("SKILL.md", "---\nname: probe\nversion: 2.0.0\n---\n")
+        write("README.md", "# probe\n")
+        write("LICENSE", "MIT\n")
+        write("CHANGELOG.md", "## v1.0.0\n")
+        try:
+            from dev_engine import check_publish_readiness
+            readiness = check_publish_readiness(str(root))
+            codes = {item["code"] for item in readiness["issues"]}
+            observed = "FAIL" if not readiness["ok"] else "PASS"
+            passed = not readiness["ok"] and "version-mismatch" in codes
+            detail = "seeded package/SKILL version mismatch"
+        except Exception as exc:  # noqa: BLE001
+            observed = "error"
+            passed = False
+            detail = str(exc)
+        record("publish-check", "seeded version mismatch", "FAIL",
+               observed, passed, detail)
+    return probes
+
+def _self_test_version_check(root, mode):
+    if mode == "installed":
+        skill = root / "SKILL.md"
+        version = _frontmatter_version(_read_text(skill)) if skill.is_file() else None
+        if not version:
+            return "FAIL", [{"code": "missing-version", "path": "SKILL.md",
+                             "detail": "SKILL.md has no version field"}], \
+                "restore a valid SKILL.md version"
+        return "PASS", [], None
+
+    versions = {}
+    try:
+        package = json.loads(_read_text(root / "package.json"))
+        versions["package.json"] = package.get("version")
+    except Exception as exc:  # noqa: BLE001
+        return "FAIL", [{"code": "invalid-package-json", "path": "package.json",
+                         "detail": str(exc)}], "fix package.json"
+    for rel in ("SKILL.md", "CHANGELOG.md", "server.json"):
+        target = root / rel
+        if not target.is_file():
+            versions[rel] = None
+            continue
+        if rel == "server.json":
+            try:
+                versions[rel] = json.loads(_read_text(target)).get("version")
+            except Exception:  # noqa: BLE001
+                versions[rel] = None
+        elif rel == "CHANGELOG.md":
+            match = re.search(r"(?m)^##\s+v?(\d+\.\d+\.\d+)", _read_text(target))
+            versions[rel] = match.group(1) if match else None
+        else:
+            versions[rel] = _frontmatter_version(_read_text(target))
+    engine_path = root / "scripts" / "dev_engine.py"
+    if engine_path.is_file():
+        match = re.search(r'(?m)^VERSION\s*=\s*["\']([^"\']+)["\']',
+                          _read_text(engine_path))
+        versions["engine"] = match.group(1) if match else None
+    else:
+        versions["engine"] = None
+    missing = sorted(key for key, value in versions.items() if not value)
+    if missing:
+        return "FAIL", [{
+            "code": "missing-version", "path": missing[0],
+            "detail": "version is missing from: %s" % ", ".join(missing),
+        }], "restore version alignment"
+    values = {value for value in versions.values() if value}
+    if len(values) != 1:
+        return "FAIL", [{
+            "code": "version-mismatch", "path": "package.json",
+            "detail": json.dumps(versions, ensure_ascii=False, sort_keys=True),
+        }], "align package, SKILL, CHANGELOG, server and engine versions"
+    return "PASS", [], None
+
+def _self_test_tool_contracts(root):
+    protocol_path = root / "scripts" / "yotta_dev_mcp.py"
+    engine_path = root / "scripts" / "dev_engine.py"
+    if not protocol_path.is_file() or not engine_path.is_file():
+        return "FAIL", [{
+            "code": "missing-protocol-source", "path": "scripts",
+            "detail": "protocol or engine source is missing",
+        }], "restore the protocol and engine sources"
+    try:
+        contracts = _protocol_tool_contracts(_read_text(protocol_path))
+        dispatch_names = _dispatch_tool_names(_read_text(engine_path))
+    except (OSError, ValueError, SyntaxError) as exc:
+        return "FAIL", [{
+            "code": "tool-contract-parse-error", "path": "scripts",
+            "detail": str(exc),
+        }], "fix the protocol or engine source"
+    protocol_names = [item["name"] for item in contracts]
+    evidence = []
+    if len(protocol_names) != len(set(protocol_names)):
+        evidence.append({"code": "tool-name-duplicate", "path": "scripts/yotta_dev_mcp.py",
+                         "detail": "duplicate tool names in mcp_tools()"})
+    if set(protocol_names) != set(dispatch_names):
+        evidence.append({
+            "code": "tool-name-drift", "path": "scripts/yotta_dev_mcp.py",
+            "detail": "protocol=%s dispatch=%s" % (
+                ",".join(sorted(protocol_names)), ",".join(sorted(dispatch_names)),
+            ),
+        })
+    drifted_schema = sorted(
+        item["name"] for item in contracts if not item["additional_properties"]
+    )
+    if drifted_schema:
+        evidence.append({
+            "code": "tool-schema-drift", "path": "scripts/yotta_dev_mcp.py",
+            "detail": "inputSchema.additionalProperties is not false for: %s"
+                      % ", ".join(drifted_schema),
+        })
+    if evidence:
+        return "FAIL", evidence, "align the protocol schemas with the engine handlers"
+    return "PASS", [], None
+
+def _self_test_write_gates(root):
+    engine_path = root / "scripts" / "dev_engine.py"
+    if not engine_path.is_file():
+        return "FAIL", [{"code": "missing-engine-source", "path": "scripts/dev_engine.py",
+                         "detail": "engine source is missing"}], "restore the engine source"
+    text = _read_text(engine_path)
+    evidence = []
+    for function_name, parameter in VERIFY_WRITE_GATES:
+        found, value = _function_default(text, function_name, parameter)
+        if not found:
+            evidence.append({
+                "code": "write-gate-missing",
+                "path": "scripts/dev_engine.py",
+                "detail": "%s(%s=...) was not found" % (function_name, parameter),
+            })
+        elif value is not False:
+            evidence.append({
+                "code": "write-gate-drift",
+                "path": "scripts/dev_engine.py",
+                "detail": "%s.%s default is %r, expected False"
+                          % (function_name, parameter, value),
+            })
+    if evidence:
+        return "FAIL", evidence, "restore fail-closed defaults for write and execute gates"
+    return "PASS", [], None
+
+def _self_test_detect_test_kind(root):
+    package_path = root / "package.json"
+    if package_path.is_file():
+        try:
+            package = json.loads(_read_text(package_path))
+            scripts = package.get("scripts") or {}
+            if isinstance(scripts, dict) and scripts.get("test"):
+                return "npm-test"
+        except Exception:  # noqa: BLE001
+            pass
+    if list(root.glob("test*.py")) or list(root.glob("tests/test*.py")):
+        return "python-unittest"
+    return None
+
+def self_test(path, mode="auto", allow_execute=False, timeout=120):
+    """Run deterministic integrity and counterexample checks on yotta-dev-mcp itself."""
+    root = Path(path)
+    if not root.exists():
+        raise ValueError("路径不存在: %s" % path)
+    if not root.is_dir():
+        raise ValueError("self_test 需要目录: %s" % path)
+    if mode not in ("auto", "source", "installed"):
+        raise ValueError("mode 必须是 auto、source 或 installed")
+    if isinstance(timeout, bool) or not isinstance(timeout, int) or not 1 <= timeout <= 600:
+        raise ValueError("timeout 必须是 1 到 600 之间的整数")
+    if mode == "auto":
+        if ((root / "package.json").is_file()
+                or (root / "scripts" / "yotta_dev_mcp.py").is_file()):
+            mode = "source"
+        elif (root / "SKILL.md").is_file():
+            mode = "installed"
+        else:
+            mode = "source"
+
+    if mode == "source":
+        required = list(VERIFY_REQUIRED_SOURCE_FILES)
+    else:
+        required = list(VERIFY_REQUIRED_INSTALLED_FILES)
+        # Lean distribution strips assets/ by platform policy; only require
+        # the banner when the copy is not lean or it already ships assets/.
+        if not _is_lean_install(root) or (root / "assets").is_dir():
+            required.extend(VERIFY_REQUIRED_INSTALLED_ASSETS)
+    missing = [rel for rel in required if not (root / rel).is_file()]
+    files_check = {
+        "id": "files",
+        "status": "FAIL" if missing else "PASS",
+        "severity": "high",
+        "claim": "required %s files are present" % mode,
+        "evidence": [{
+            "code": "missing-file", "path": rel,
+            "detail": "required file is missing",
+        } for rel in missing],
+        "next_step": "restore the missing files" if missing else None,
+    }
+
+    version_status, version_evidence, version_next = _self_test_version_check(root, mode)
+    version_check = {
+        "id": "versions" if mode == "source" else "skill-version",
+        "status": version_status,
+        "severity": "high",
+        "claim": ("package, SKILL, CHANGELOG, server and engine versions align"
+                  if mode == "source" else "SKILL.md declares a version"),
+        "evidence": version_evidence,
+        "next_step": version_next,
+    }
+
+    checks = [files_check, version_check]
+    unverified = []
+    if mode == "source":
+        tool_status, tool_evidence, tool_next = _self_test_tool_contracts(root)
+        checks.append({
+            "id": "tool-contracts",
+            "status": tool_status,
+            "severity": "high",
+            "claim": "protocol tool schemas match the engine dispatch handlers",
+            "evidence": tool_evidence,
+            "next_step": tool_next,
+        })
+        gate_status, gate_evidence, gate_next = _self_test_write_gates(root)
+        checks.append({
+            "id": "write-gates",
+            "status": gate_status,
+            "severity": "high",
+            "claim": "write and execute gates default to fail-closed",
+            "evidence": gate_evidence,
+            "next_step": gate_next,
+        })
+    else:
+        unverified.extend([
+            {"level": "L0", "claim": "protocol tool schemas match engine handlers",
+             "status": "UNVERIFIED", "reason": "installed mode has no protocol source",
+             "required": False, "next_step": "run self_test on the source checkout"},
+            {"level": "L0", "claim": "write and execute gates default to fail-closed",
+             "status": "UNVERIFIED", "reason": "installed mode has no engine source",
+             "required": False, "next_step": "run self_test on the source checkout"},
+        ])
+
+    probes = _self_test_counterexamples()
+    probe_failed = [item for item in probes if not item["passed"]]
+    checks.append({
+        "id": "verifier-counterexamples",
+        "status": "FAIL" if probe_failed else "PASS",
+        "severity": "high",
+        "claim": "seeded defects and mutation controls turn the verifier red or unknown",
+        "evidence": probes,
+        "next_step": "fix the verifier so every counterexample is detected"
+                     if probe_failed else None,
+    })
+
+    if allow_execute:
+        kind = _self_test_detect_test_kind(root)
+        if kind is None:
+            checks.append({
+                "id": "tests",
+                "status": "UNKNOWN",
+                "severity": "medium",
+                "claim": "the project test suite passes",
+                "evidence": [{"code": "test-runner-missing", "path": ".",
+                              "detail": "no whitelisted test runner was detected"}],
+                "next_step": "declare a test script or add test_*.py files",
+            })
+        else:
+            try:
+                from dev_engine import run_checks
+                result = run_checks(kind, str(root), timeout=timeout,
+                                    allow_execute=True)
+                output = result.get("output") or ""
+                checks.append({
+                    "id": "tests",
+                    "status": "PASS" if result.get("passed") else "FAIL",
+                    "severity": "high",
+                    "claim": "the project test suite passes",
+                    "evidence": [{"code": "test-run", "path": ".",
+                                  "detail": result.get("summary") or "test run finished"}],
+                    "next_step": None if result.get("passed")
+                                else "fix the failing tests and rerun self_test",
+                    "command": {
+                        "kind": kind,
+                        "cwd": ".",
+                        "exit_code": result.get("exit_code"),
+                        "output_hash": hashlib.sha256(
+                            output.encode("utf-8")
+                        ).hexdigest(),
+                        "timed_out": bool(result.get("timed_out")),
+                    },
+                })
+            except Exception as exc:  # noqa: BLE001
+                checks.append({
+                    "id": "tests",
+                    "status": "FAIL",
+                    "severity": "high",
+                    "claim": "the project test suite passes",
+                    "evidence": [{"code": "test-run-error", "path": ".",
+                                  "detail": str(exc)}],
+                    "next_step": "fix the local test runner",
+                })
+    else:
+        unverified.append({
+            "level": "L2",
+            "claim": "the project test suite passes",
+            "status": "UNVERIFIED",
+            "reason": "allow_execute=false",
+            "required": False,
+            "next_step": "rerun self_test with allow_execute=true",
+        })
+
+    failed = any(item["status"] == "FAIL" for item in checks)
+    unknown = any(item["status"] == "UNKNOWN" for item in checks)
+    if failed:
+        status = "FAIL"
+    elif unknown:
+        status = "UNKNOWN"
+    else:
+        status = "PASS"
+    counts = {
+        "passed": sum(1 for item in checks if item["status"] == "PASS"),
+        "failed": sum(1 for item in checks if item["status"] == "FAIL"),
+        "unknown": sum(1 for item in checks if item["status"] == "UNKNOWN"),
+        "unverified": len(unverified),
+    }
+    evidence = []
+    for item in checks:
+        evidence.extend(item.get("evidence") or [])
+    return {
+        "status": status,
+        "root": str(root.resolve()),
+        "mode": mode,
+        "checks": checks,
+        "unverified_claims": unverified,
+        "summary": counts,
+        "evidence": evidence[:EVIDENCE_LIMIT],
+    }