@trac3er/oh-my-god 2.1.9 → 2.2.2

package/runtime/issue_surface.py

@@ -0,0 +1,362 @@
+from __future__ import annotations
+
+from dataclasses import asdict, dataclass
+from datetime import datetime, timezone
+import json
+import os
+from pathlib import Path
+from typing import cast
+
+from runtime.forge_agents import collect_forge_evidence_issues
+from runtime.incident_replay import collect_incident_signals
+from runtime.plugin_diagnostics import run_plugin_diagnostics
+from runtime.plugin_interop import conflict_severity_to_issue_severity
+from runtime.session_health import collect_session_health_risks
+from tools.session_snapshot import collect_snapshot_signals
+
+
+_ALL_SURFACES: tuple[str, ...] = (
+    "live_session",
+    "forge_runs",
+    "hooks",
+    "skills",
+    "mcps",
+    "plugin_interop",
+    "governed_tools",
+    "domain_pipelines",
+)
+
+_SEVERITY_RANK: dict[str, int] = {
+    "critical": 0,
+    "high": 1,
+    "medium": 2,
+    "low": 3,
+    "info": 4,
+}
+
+
+@dataclass(slots=True)
+class Issue:
+    id: str
+    severity: str
+    surface: str
+    title: str
+    description: str
+    fix_guidance: str
+    evidence_links: list[str]
+    approval_required: bool
+    approval_reason: str
+
+    def to_dict(self) -> dict[str, object]:
+        return asdict(self)
+
+
+@dataclass(slots=True)
+class IssueReport:
+    run_id: str
+    timestamp: str
+    issues: list[Issue]
+    summary: dict[str, object]
+
+    def to_dict(self) -> dict[str, object]:
+        return {
+            "schema": "IssueReport",
+            "run_id": self.run_id,
+            "timestamp": self.timestamp,
+            "issues": [issue.to_dict() for issue in self.issues],
+            "summary": dict(self.summary),
+        }
+
+
+class IssueSurface:
+    def __init__(self, project_dir: str | None = None) -> None:
+        self.project_dir: str = str(Path(project_dir or os.environ.get("CLAUDE_PROJECT_DIR", os.getcwd())).resolve())
+
+    def scan(self, run_id: str, surfaces: list[str] | tuple[str, ...] | None = None) -> IssueReport:
+        selected_surfaces = self._normalize_surfaces(surfaces)
+        issues: list[Issue] = []
+        scan_map = {
+            "live_session": self._scan_live_session,
+            "forge_runs": self._scan_forge_runs,
+            "hooks": self._scan_hooks,
+            "skills": self._scan_skills,
+            "mcps": self._scan_mcps,
+            "plugin_interop": self._scan_plugin_interop,
+            "governed_tools": self._scan_governed_tools,
+            "domain_pipelines": self._scan_domain_pipelines,
+        }
+
+        for surface in selected_surfaces:
+            scanner = scan_map.get(surface)
+            if scanner is not None:
+                issues.extend(scanner(run_id))
+
+        ranked_issues = sorted(
+            issues,
+            key=lambda issue: (
+                _SEVERITY_RANK.get(issue.severity, 99),
+                issue.surface,
+                issue.id,
+            ),
+        )
+        summary = self._build_summary(ranked_issues, selected_surfaces)
+        report = IssueReport(
+            run_id=run_id,
+            timestamp=datetime.now(timezone.utc).isoformat(),
+            issues=ranked_issues,
+            summary=summary,
+        )
+        report_path = self.emit_report(run_id, report)
+        report.summary["report_path"] = report_path
+        return report
+
+    def simulate_failure(self, surface: str, scenario: str) -> dict[str, object]:
+        normalized_surface = str(surface).strip().lower() or "unknown"
+        normalized_scenario = str(scenario).strip().lower() or "unknown"
+        dangerous_probe = any(
+            token in normalized_scenario
+            for token in ("delete", "overwrite", "rm -rf", "exfiltrate", "credential", "chmod")
+        )
+        severity = "high" if dangerous_probe else "medium"
+        approval_required = dangerous_probe
+        issue = Issue(
+            id=f"sim-{normalized_surface}-{abs(hash(normalized_scenario)) % 100000}",
+            severity=severity,
+            surface=normalized_surface,
+            title="Sandboxed red-team simulation result",
+            description=(
+                "Probe classified as potentially destructive and blocked in simulation mode."
+                if dangerous_probe
+                else "Probe classified as non-destructive simulation scenario."
+            ),
+            fix_guidance=(
+                "Keep simulation read-only; require signed approval artifact before any real mutation workflow."
+                if dangerous_probe
+                else "Document scenario and add replay fixture for continued diagnostics."
+            ),
+            evidence_links=[f"sim://{normalized_surface}/{normalized_scenario}"],
+            approval_required=approval_required,
+            approval_reason=(
+                "signed approval required before leaving read-only simulation mode"
+                if dangerous_probe
+                else ""
+            ),
+        )
+        return {
+            "schema": "IssueSimulationResult",
+            "surface": normalized_surface,
+            "scenario": normalized_scenario,
+            "mutated": False,
+            "issue": issue.to_dict(),
+        }
+
+    def emit_report(self, run_id: str, report: IssueReport) -> str:
+        evidence_dir = Path(self.project_dir) / ".omg" / "evidence" / "issues"
+        evidence_dir.mkdir(parents=True, exist_ok=True)
+        path = evidence_dir / f"{run_id}.json"
+        tmp_path = path.with_name(f"{path.name}.tmp")
+        _ = tmp_path.write_text(json.dumps(report.to_dict(), indent=2, ensure_ascii=True) + "\n", encoding="utf-8")
+        os.replace(tmp_path, path)
+        return str(path.relative_to(self.project_dir)).replace("\\", "/")
+
+    def _normalize_surfaces(self, surfaces: list[str] | tuple[str, ...] | None) -> list[str]:
+        if not surfaces:
+            return list(_ALL_SURFACES)
+        normalized: list[str] = []
+        for item in surfaces:
+            candidate = str(item).strip().lower()
+            if candidate in _ALL_SURFACES and candidate not in normalized:
+                normalized.append(candidate)
+        return normalized or list(_ALL_SURFACES)
+
+    def _build_summary(self, issues: list[Issue], surfaces: list[str]) -> dict[str, object]:
+        by_severity = {name: 0 for name in _SEVERITY_RANK}
+        by_surface = {surface: 0 for surface in surfaces}
+        for issue in issues:
+            if issue.severity in by_severity:
+                by_severity[issue.severity] += 1
+            by_surface[issue.surface] = by_surface.get(issue.surface, 0) + 1
+        return {
+            "total": len(issues),
+            "surfaces_scanned": surfaces,
+            "by_severity": by_severity,
+            "by_surface": by_surface,
+            "requires_signed_approval": any(issue.approval_required for issue in issues),
+        }
+
+    def _issue_from_payload(self, run_id: str, idx: int, payload: dict[str, object]) -> Issue:
+        severity = str(payload.get("severity", "info"))
+        approval_required = bool(payload.get("approval_required", severity in {"critical", "high"}))
+        approval_reason = str(payload.get("approval_reason", "")).strip()
+        if approval_required and not approval_reason:
+            approval_reason = "signed approval required before fix execution"
+        surface = str(payload.get("surface", "unknown"))
+        return Issue(
+            id=f"{run_id}-{surface}-{idx}",
+            severity=severity,
+            surface=surface,
+            title=str(payload.get("title", "Issue detected")),
+            description=str(payload.get("description", "")),
+            fix_guidance=str(payload.get("fix_guidance", "")),
+            evidence_links=self._to_str_list(payload.get("evidence_links")),
+            approval_required=approval_required,
+            approval_reason=approval_reason,
+        )
+
+    def _to_str_list(self, value: object) -> list[str]:
+        if not isinstance(value, list):
+            return []
+        result: list[str] = []
+        for item in cast(list[object], value):
+            text = str(item).strip()
+            if text:
+                result.append(text)
+        return result
+
+    def _scan_live_session(self, run_id: str) -> list[Issue]:
+        payloads = collect_session_health_risks(self.project_dir, run_id)
+        snapshots = collect_snapshot_signals(str(Path(self.project_dir) / ".omg" / "state"))
+        issues = [self._issue_from_payload(run_id, idx, payload) for idx, payload in enumerate(payloads, start=1)]
+        if int(snapshots.get("snapshot_count", 0)) == 0:
+            issues.append(
+                self._issue_from_payload(
+                    run_id,
+                    len(issues) + 1,
+                    {
+                        "severity": "low",
+                        "surface": "live_session",
+                        "title": "No session snapshots available",
+                        "description": "Issue scan found no state snapshots for replay diagnostics.",
+                        "fix_guidance": "Enable snapshot capture or run a snapshot before deep issue triage.",
+                        "evidence_links": [".omg/state/snapshots/"],
+                        "approval_required": False,
+                        "approval_reason": "",
+                    },
+                )
+            )
+        return issues
+
+    def _scan_forge_runs(self, run_id: str) -> list[Issue]:
+        payloads = collect_forge_evidence_issues(self.project_dir, run_id)
+        return [self._issue_from_payload(run_id, idx, payload) for idx, payload in enumerate(payloads, start=1)]
+
+    def _scan_hooks(self, run_id: str) -> list[Issue]:
+        signals = collect_incident_signals(self.project_dir, run_id)
+        payloads = signals.get("hook_issues", [])
+        if not isinstance(payloads, list):
+            return []
+        typed_payloads = [payload for payload in payloads if isinstance(payload, dict)]
+        return [
+            self._issue_from_payload(run_id, idx, payload)
+            for idx, payload in enumerate(typed_payloads, start=1)
+        ]
+
+    def _scan_skills(self, run_id: str) -> list[Issue]:
+        diagnostics = run_plugin_diagnostics(root=self.project_dir, live=False)
+        records = diagnostics.get("records", [])
+        if not isinstance(records, list):
+            return []
+        typed_records = [record for record in records if isinstance(record, dict)]
+        skill_records = [record for record in typed_records if record.get("source") == "skill_registry"]
+        if skill_records:
+            return []
+        return [
+            self._issue_from_payload(
+                run_id,
+                1,
+                {
+                    "severity": "medium",
+                    "surface": "skills",
+                    "title": "No compiled skills discovered",
+                    "description": "Diagnostic scan did not detect skill-registry entries for the active workspace.",
+                    "fix_guidance": "Regenerate skill registry artifacts and verify .omg/state/skill_registry/compact.json.",
+                    "evidence_links": [".omg/state/skill_registry/compact.json"],
+                    "approval_required": False,
+                    "approval_reason": "",
+                },
+            )
+        ]
+
+    def _scan_mcps(self, run_id: str) -> list[Issue]:
+        diagnostics = run_plugin_diagnostics(root=self.project_dir, live=False)
+        records = diagnostics.get("records", [])
+        if not isinstance(records, list):
+            return []
+        typed_records = [record for record in records if isinstance(record, dict)]
+        mcp_records = [record for record in typed_records if record.get("mcp_servers")]
+        if mcp_records:
+            return []
+        return [
+            self._issue_from_payload(
+                run_id,
+                1,
+                {
+                    "severity": "medium",
+                    "surface": "mcps",
+                    "title": "No MCP servers discovered",
+                    "description": "Issue scan found no configured MCP servers across discovered host configs.",
+                    "fix_guidance": "Register required MCP servers and rerun diagnostics.",
+                    "evidence_links": [".mcp.json", ".claude-plugin/mcp.json"],
+                    "approval_required": False,
+                    "approval_reason": "",
+                },
+            )
+        ]
+
+    def _scan_plugin_interop(self, run_id: str) -> list[Issue]:
+        diagnostics = run_plugin_diagnostics(root=self.project_dir, live=False)
+        conflicts = diagnostics.get("conflicts", [])
+        if not isinstance(conflicts, list):
+            return []
+
+        issues: list[Issue] = []
+        typed_conflicts = [conflict for conflict in conflicts if isinstance(conflict, dict)]
+        for idx, conflict in enumerate(typed_conflicts, start=1):
+            severity = conflict_severity_to_issue_severity(str(conflict.get("severity", "info")))
+            code = str(conflict.get("code", "conflict"))
+            detail = str(conflict.get("detail", ""))
+            fix_guidance = str(conflict.get("next_action", ""))
+            approval_required = severity in {"critical", "high"}
+            issues.append(
+                self._issue_from_payload(
+                    run_id,
+                    idx,
+                    {
+                        "severity": severity,
+                        "surface": "plugin_interop",
+                        "title": f"Plugin interop conflict: {code}",
+                        "description": detail,
+                        "fix_guidance": fix_guidance,
+                        "evidence_links": [".omg/state/plugins-allowlist.yaml", ".mcp.json"],
+                        "approval_required": approval_required,
+                        "approval_reason": (
+                            "signed approval required to modify plugin ownership on blocker/high conflicts"
+                            if approval_required
+                            else ""
+                        ),
+                    },
+                )
+            )
+        return issues
+
+    def _scan_governed_tools(self, run_id: str) -> list[Issue]:
+        signals = collect_incident_signals(self.project_dir, run_id)
+        payloads = signals.get("governed_tool_issues", [])
+        if not isinstance(payloads, list):
+            return []
+        typed_payloads = [payload for payload in payloads if isinstance(payload, dict)]
+        return [
+            self._issue_from_payload(run_id, idx, payload)
+            for idx, payload in enumerate(typed_payloads, start=1)
+        ]
+
+    def _scan_domain_pipelines(self, run_id: str) -> list[Issue]:
+        payloads = collect_forge_evidence_issues(self.project_dir, run_id, domain_pipeline_only=True)
+        return [self._issue_from_payload(run_id, idx, payload) for idx, payload in enumerate(payloads, start=1)]
+
+
+__all__ = [
+    "Issue",
+    "IssueReport",
+    "IssueSurface",
+]

package/runtime/merge_writer.py

@@ -0,0 +1,320 @@
+"""Merge-writer lock — single authorized writer for main-tree mutations.
+
+Enforces one-writer/many-readers semantics:
+- Read-only workers run concurrently without obtaining the merge lock.
+- Mutation-capable workers MUST hold the merge-writer lock AND run
+  inside a worktree before applying changes back to the main tree.
+
+Lock file:     .omg/state/merge-writer.lock   (JSON)
+Provenance:    .omg/evidence/merge-writer-<run_id>.json
+"""
+from __future__ import annotations
+
+import json
+import os
+from dataclasses import dataclass
+from datetime import datetime, timezone
+from pathlib import Path
+from typing import Any
+from typing import cast
+
+
+# ---------------------------------------------------------------------------
+# Exceptions
+# ---------------------------------------------------------------------------
+
+class MergeWriterAuthorizationError(RuntimeError):
+    """Raised when an unauthorized worker attempts a main-tree mutation."""
+
+    def __init__(self, run_id: str, reason: str) -> None:
+        self.run_id = run_id
+        self.reason = reason
+        super().__init__(
+            f"merge-writer authorization denied for run_id={run_id}: {reason}"
+        )
+
+
+# ---------------------------------------------------------------------------
+# Token dataclass
+# ---------------------------------------------------------------------------
+
+@dataclass(frozen=True)
+class MergeWriterToken:
+    """Proof-of-ownership token returned by MergeWriter.acquire()."""
+
+    run_id: str
+    acquired_at: str
+    lock_path: str
+    provenance_path: str
+
+
+# ---------------------------------------------------------------------------
+# MergeWriter
+# ---------------------------------------------------------------------------
+
+def _now_iso() -> str:
+    return datetime.now(timezone.utc).isoformat()
+
+
+def _write_atomic_json(path: Path, payload: dict[str, Any]) -> None:
+    path.parent.mkdir(parents=True, exist_ok=True)
+    tmp_path = path.with_name(f"{path.name}.tmp")
+    tmp_path.write_text(
+        json.dumps(payload, indent=2, ensure_ascii=True) + "\n",
+        encoding="utf-8",
+    )
+    os.replace(tmp_path, path)
+
+
+def _read_json(path: Path) -> dict[str, Any]:
+    if not path.exists():
+        return {}
+    try:
+        payload = json.loads(path.read_text(encoding="utf-8"))
+    except (OSError, json.JSONDecodeError):
+        return {}
+    return payload if isinstance(payload, dict) else {}
+
+
+class MergeWriter:
+    """Single merge-writer lock manager with provenance.
+
+    Usage::
+
+        mw = MergeWriter("/path/to/project")
+        token = mw.acquire("run-abc", reason="release merge")
+        # ... perform mutation ...
+        mw.release(token)
+    """
+
+    def __init__(self, project_dir: str | None = None) -> None:
+        if project_dir:
+            self.project_dir = Path(project_dir).resolve()
+        else:
+            self.project_dir = Path(
+                os.environ.get("CLAUDE_PROJECT_DIR", os.getcwd())
+            ).resolve()
+
+    # --- Path helpers -------------------------------------------------------
+
+    def _lock_path(self) -> Path:
+        return self.project_dir / ".omg" / "state" / "merge-writer.lock"
+
+    def _provenance_path(self, run_id: str) -> Path:
+        return self.project_dir / ".omg" / "evidence" / f"merge-writer-{run_id}.json"
+
+    def _active_coordinator_run_id(self) -> str:
+        shadow_path = self.project_dir / ".omg" / "shadow" / "active-run"
+        if shadow_path.exists():
+            try:
+                shadow = shadow_path.read_text(encoding="utf-8").strip()
+            except OSError:
+                shadow = ""
+            if shadow:
+                return shadow
+
+        state_dir = self.project_dir / ".omg" / "state" / "release_run_coordinator"
+        if not state_dir.exists():
+            return ""
+        candidates = sorted(state_dir.glob("*.json"), key=lambda p: p.stat().st_mtime)
+        for path in reversed(candidates):
+            payload = _read_json(path)
+            if not payload:
+                continue
+            payload_dict = cast(dict[str, object], payload)
+            run_id = str(payload_dict.get("run_id", "")).strip()
+            if run_id:
+                return run_id
+        return ""
+
+    # --- Public API ---------------------------------------------------------
+
+    def acquire(self, run_id: str, reason: str = "") -> MergeWriterToken:
+        """Acquire the merge-writer lock.
+
+        Args:
+            run_id:  Unique identifier for the writer (typically a release run).
+            reason:  Human-readable reason for acquiring the lock.
+
+        Returns:
+            A ``MergeWriterToken`` proving ownership.
+
+        Raises:
+            MergeWriterAuthorizationError: If the lock is already held by
+                a different run_id.
+        """
+        lock_path = self._lock_path()
+        existing = _read_json(lock_path)
+
+        if existing:
+            existing_run_id = existing.get("run_id", "")
+            if existing_run_id and existing_run_id != run_id:
+                raise MergeWriterAuthorizationError(
+                    run_id,
+                    f"lock already held by run_id={existing_run_id}",
+                )
+
+        now = _now_iso()
+        lock_payload: dict[str, Any] = {
+            "run_id": run_id,
+            "acquired_at": now,
+            "reason": reason,
+            "pid": os.getpid(),
+        }
+        _write_atomic_json(lock_path, lock_payload)
+
+        provenance_path = self._provenance_path(run_id)
+        rel_lock = str(lock_path.relative_to(self.project_dir)).replace("\\", "/")
+        rel_prov = str(provenance_path.relative_to(self.project_dir)).replace("\\", "/")
+
+        return MergeWriterToken(
+            run_id=run_id,
+            acquired_at=now,
+            lock_path=rel_lock,
+            provenance_path=rel_prov,
+        )
+
+    def release(self, token: MergeWriterToken) -> dict[str, Any]:
+        """Release the merge-writer lock and write provenance.
+
+        Args:
+            token: The token returned by :meth:`acquire`.
+
+        Returns:
+            Provenance evidence dict written to disk.
+
+        Raises:
+            MergeWriterAuthorizationError: If the caller does not own the lock.
+        """
+        lock_path = self._lock_path()
+        existing = _read_json(lock_path)
+
+        if existing.get("run_id") != token.run_id:
+            raise MergeWriterAuthorizationError(
+                token.run_id,
+                "cannot release: lock is not held by this run_id",
+            )
+
+        provenance: dict[str, Any] = {
+            "schema": "MergeWriterProvenance",
+            "run_id": token.run_id,
+            "acquired_at": token.acquired_at,
+            "released_at": _now_iso(),
+            "lock_path": token.lock_path,
+            "provenance_path": token.provenance_path,
+            "pid": existing.get("pid"),
+            "reason": existing.get("reason", ""),
+        }
+
+        provenance_path = self._provenance_path(token.run_id)
+        _write_atomic_json(provenance_path, provenance)
+
+        # Remove lock file
+        try:
+            lock_path.unlink(missing_ok=True)
+        except OSError:
+            pass
+
+        return provenance
+
+    def is_locked(self) -> bool:
+        """Return True if the merge-writer lock is currently held."""
+        lock_path = self._lock_path()
+        if not lock_path.exists():
+            return False
+        existing = _read_json(lock_path)
+        return bool(existing.get("run_id"))
+
+    def get_current_owner(self) -> str | None:
+        """Return run_id of the current lock holder, or None."""
+        existing = _read_json(self._lock_path())
+        run_id = existing.get("run_id", "")
+        return run_id if run_id else None
+
+    def check_authorization_details(self, run_id: str) -> dict[str, Any]:
+        owner_run_id = self.get_current_owner() or ""
+        active_run_id = self._active_coordinator_run_id()
+        owner_matches = bool(owner_run_id) and owner_run_id == run_id
+        active_matches = (not active_run_id) or active_run_id == run_id
+        authorized = owner_matches and active_matches
+
+        reason = "authorized"
+        if not owner_run_id:
+            reason = "merge_writer_lock_missing"
+        elif owner_run_id != run_id:
+            reason = f"merge_writer_owner_mismatch:{owner_run_id}"
+        elif active_run_id and active_run_id != run_id:
+            reason = f"active_run_mismatch:{active_run_id}"
+
+        return {
+            "run_id": run_id,
+            "owner_run_id": owner_run_id,
+            "active_run_id": active_run_id,
+            "owner_matches": owner_matches,
+            "active_matches": active_matches,
+            "authorized": authorized,
+            "reason": reason,
+        }
+
+    def check_authorization(self, run_id: str) -> bool:
+        """Return True if *run_id* is the current authorized writer."""
+        return bool(self.check_authorization_details(run_id).get("authorized"))
+
+    # --- Governance helpers -------------------------------------------------
+
+    def require_authorization(
+        self,
+        run_id: str,
+        *,
+        mutation_type: str = "merge",
+        isolation: str = "none",
+    ) -> None:
+        """Gate a mutation attempt: block if unauthorized.
+
+        Mutation-capable workers MUST:
+        1. Run inside a worktree (isolation == "worktree").
+        2. Hold the merge-writer lock for their run_id.
+
+        Read-only workers (isolation != "worktree") are allowed through
+        without the lock.
+
+        Raises:
+            MergeWriterAuthorizationError: On unauthorized mutation.
+        """
+        normalized_isolation = str(isolation).strip().lower()
+
+        # Read-only workers never need the merge-writer lock
+        if normalized_isolation != "worktree":
+            return
+
+        if not self.is_locked():
+            raise MergeWriterAuthorizationError(
+                run_id,
+                f"mutation_type={mutation_type} requires merge-writer lock "
+                f"but no lock is held",
+            )
+
+        details = self.check_authorization_details(run_id)
+        if not details.get("authorized"):
+            owner = str(details.get("owner_run_id") or "unknown")
+            active = str(details.get("active_run_id") or "")
+            if active and active != run_id:
+                raise MergeWriterAuthorizationError(
+                    run_id,
+                    f"mutation_type={mutation_type} denied: active coordinator run_id="
+                    f"{active}, not {run_id}",
+                )
+            raise MergeWriterAuthorizationError(
+                run_id,
+                f"mutation_type={mutation_type} denied: lock held by "
+                f"run_id={owner}, not {run_id}",
+            )
+
+
+# ---------------------------------------------------------------------------
+# Module-level factory
+# ---------------------------------------------------------------------------
+
+def get_merge_writer(project_dir: str | None = None) -> MergeWriter:
+    """Factory for MergeWriter instances."""
+    return MergeWriter(project_dir)