feat: make crosswalk and stop-line offsets solvable

Step 1 of the control-marking task, and deliberately server-only: no handle is
drawn yet. This project already shipped a range handle for `profile.interval`,
which compileGeometry ignores, so the control dragged and changed nothing. The
consumer comes first now.

Two kinds join the taxonomy — `junction-crosswalk-inset` and
`junction-stop-line-offset`, both on the existing `junction-approach` anchor. The
solver writes them onto the approach entry, `applyDirectJunctionPlans` carries
them onto the compiled approach, and `compileControlMarkings` reads them in place
of the module constants it used for every junction. They move markings without
reshaping the junction, so unlike width and cutback they deliberately do not
trigger a boundary recompute.

`applyJunctionConstraint` becomes an explicit switch. Its trailing `else` had
meant every kind that was not approach-width fell through to the cutback
validator, so a new kind would have been silently validated and written as a
cutback. The same non-exhaustive shape in the test fixture's `valueFor` is fixed
the same way, and now throws for an unnamed kind rather than answering with a
corner radius.

design.md's taxonomy is updated with it — a test asserts the two cannot drift,
which is what caught the omission.

Measured on a 41-road workspace with 8 crossings: both constraints change their
marking geometry, neither drags the other, and out-of-range blocks instead of
clamping. That measurement is not in the suite: the synthetic junction resolves
`junction_inset_m` to 0 because its crossing never binds to a plan, and the
committed OSM fixture has no crossings at all. The tests assert the wiring the
handles will depend on — values reaching the approach entry, distinct branches,
blocking diagnostics — and the gap is recorded in the test itself.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

.trellis/scripts/common/active_task.py

@@ -0,0 +1,765 @@
+#!/usr/bin/env python3
+"""Session-scoped active task resolution.
+
+The user-facing concept is a single "active task". Trellis stores that pointer
+per AI session/window under `.trellis/.runtime/sessions/`; without a stable
+session key there is no active task.
+"""
+
+from __future__ import annotations
+
+import hashlib
+import json
+import os
+import re
+import sys
+import time
+from dataclasses import dataclass
+from datetime import datetime, timezone
+from pathlib import Path
+from typing import Any
+
+DIR_WORKFLOW = ".trellis"
+DIR_TASKS = "tasks"
+DIR_RUNTIME = ".runtime"
+DIR_SESSIONS = "sessions"
+DIR_SHELL_TICKETS = "shell-tickets"
+# Pre-0.6.13 name, when the bridge was Cursor-only. Still read so a session that
+# was mid-command across an upgrade does not silently degrade; never written.
+# Tickets are 30-second ephemera, so the old directory ages out by itself —
+# there is nothing to migrate, only a glob on a directory that is normally
+# absent. The alternative (ignore it) would land its one lost command on the
+# platform that works today.
+DIR_LEGACY_CURSOR_SHELL_TICKETS = "cursor-shell"
+SHELL_TICKET_TTL_SECONDS = 30
+TASK_SESSION_COMMANDS = {"start", "current", "finish"}
+
+_SESSION_KEYS = ("session_id", "sessionId", "sessionID")
+_CONVERSATION_KEYS = ("conversation_id", "conversationId", "conversationID")
+_TRANSCRIPT_KEYS = ("transcript_path", "transcriptPath", "transcript")
+_NESTED_KEYS = ("input", "properties", "event", "hook_input", "hookInput")
+_KNOWN_PLATFORMS = {
+    "claude",
+    "codex",
+    "cursor",
+    "opencode",
+    "gemini",
+    "droid",
+    "qoder",
+    "codebuddy",
+    "kiro",
+    "copilot",
+    "pi",
+    "trae",
+    "grok",
+    "kimi",
+    "zcode",
+    "snow",
+    "dsh",
+}
+
+# Every name below records how it was checked. Do NOT add a name by analogy
+# with a neighbour: a 2026-08-05 audit of all 21 platforms found 12 of the 21
+# declared names had never existed anywhere — they were pattern-guessed from a
+# `<PLATFORM>_SESSION_ID` shape no vendor agreed to, and the uniformity was the
+# only "evidence" behind them. A platform with no verified name belongs in no
+# table; it resolves through TRELLIS_CONTEXT_ID or its hook/plugin bridge.
+_ENV_SESSION_KEYS: tuple[tuple[str, tuple[str, ...]], ...] = (
+    # REAL (reported 2026-08-13 against DSH 0.1.0-rc.6 by @SajoLuo, from a live
+    # run: DSH exports DSH_SESSION_ID plus DSH_SHELL=1 into its managed shell).
+    # MUST STAY FIRST. A DSH session can inherit an outer host's identity — a
+    # DSH launched from Codex still carries CODEX_THREAD_ID — and the untargeted
+    # lookup below walks this table in order, so any earlier entry would claim
+    # the session and write a foreign `codex_<thread>` pointer for DSH work.
+    # DSH_SESSION_ID is the only name here no other vendor sets, so first place
+    # is safe: it cannot mis-claim a non-DSH session.
+    ("dsh", ("DSH_SESSION_ID",)),
+    # REAL, undocumented (verified 2026-08-05 in a live Claude Code 2.1.221 bash
+    # child; absent from code.claude.com/docs/en/env-vars). CLAUDE_SESSION_ID
+    # was removed here — verified absent from that same live environment.
+    ("claude", ("CLAUDE_CODE_SESSION_ID",)),
+    # REAL, undocumented (verified 2026-08-05: injected by codex-cli 0.146.0
+    # into shell children, absent from the parent env; openai/codex#19937).
+    # CODEX_SESSION_ID was removed — absent from a live `codex exec` env.
+    ("codex", ("CODEX_THREAD_ID",)),
+    # REAL but HOOK-SCOPE ONLY (verified 2026-08-05): set by Gemini's
+    # hookRunner.ts. Its shell tool builds the child env in
+    # shellExecutionService.ts and adds only GEMINI_CLI/TERM/PAGER/GIT_PAGER, so
+    # this never reaches a bash child — it resolves only inside a hook process.
+    ("gemini", ("GEMINI_SESSION_ID",)),
+    # REAL but HOOK-SCOPE ONLY (verified 2026-08-05): docs.qoder.com/zh/
+    # extensions/hooks documents it as injected during hook execution by the
+    # Qoder *IDE plugin*. Absent from the Qoder CLI hook docs and from Lingma.
+    ("qoder", ("QODER_SESSION_ID",)),
+    # UNVERIFIED (2026-08-05): absent from kiro.dev/docs/hooks/, but Dynatrace
+    # dtctl, oh-my-agent and gastown all key agent detection on it and one notes
+    # it is "set in both interactive and --no-interactive". Kept because that is
+    # absence of evidence, not evidence of absence. To settle: run
+    # `env | grep KIRO` from a Kiro shell-tool call on a machine with Kiro.
+    ("kiro", ("KIRO_SESSION_ID",)),
+    # UNVERIFIED (2026-08-05): absent from docs.github.com/en/copilot/reference/
+    # hooks-reference and from the CLI programmatic reference. To settle: run
+    # `copilot help environment` (the authoritative list per those docs) — not
+    # runnable here, the CLI is not installed and copilot-cli ships no source.
+    ("copilot", ("COPILOT_SESSION_ID", "COPILOT_SESSIONID")),
+    # REASONED, UNVERIFIED (2026-08-05): ZCode is closed-source and not
+    # installable here. It mirrors Claude's naming elsewhere (CLAUDE_PLUGIN_ROOT
+    # / CLAUDE_PLUGIN_DATA compat aliases are in its docs), and the previously
+    # declared CLAUDE_SESSION_ID does not exist on Claude Code either — so the
+    # name ZCode would actually reuse is CLAUDE_CODE_SESSION_ID. Try that first,
+    # keep the historical name as a fallback: if neither exists nothing changes.
+    # Platform-scoped lookup (_iter_env_keys filters by platform name), so the
+    # entry only fires once the resolver detected "zcode" — no collision with
+    # the claude entry above.
+    ("zcode", ("CLAUDE_CODE_SESSION_ID", "CLAUDE_SESSION_ID")),
+    # REAL by vendor design (verified 2026-08-05): Snow's sessionIdentityEnv.ts
+    # exports SNOW_SESSION_ID into hook/terminal/sub-agent children and names
+    # Trellis in its source header. TRELLIS_CONTEXT_ID stays the preferred
+    # override — Snow sets that too.
+    ("snow", ("SNOW_SESSION_ID",)),
+)
+_ENV_CONVERSATION_KEYS: tuple[tuple[str, tuple[str, ...]], ...] = (
+    # REAL in cursor-agent (CLI), undocumented (verified 2026-08-05: the value
+    # matches ~/.cursor/chats/<ws>/<id>). The Cursor *IDE* is unverified — a
+    # 2026-05 forum request for it drew no staff reply. The invented
+    # CURSOR_SESSION_ID was removed from the session table: empty in a live
+    # cursor-agent shell. Cursor's other path is the shell ticket below
+    # (_lookup_shell_ticket_context_key), which is not Cursor-specific.
+    ("cursor", ("CURSOR_CONVERSATION_ID", "CURSOR_CONVERSATIONID")),
+)
+_ENV_TRANSCRIPT_KEYS: tuple[tuple[str, tuple[str, ...]], ...] = (
+    # REAL but HOOK-SCOPE ONLY (verified 2026-08-05): documented for Cursor hook
+    # scripts; empty in the agent's own shell env.
+    ("cursor", ("CURSOR_TRANSCRIPT_PATH",)),
+    # UNVERIFIED — never researched. The 2026-08-05 audit covered the session
+    # table only, so do not infer these are real *or* fake from that work
+    # (CLAUDE_/CODEX_TRANSCRIPT_PATH were removed because those two *were*
+    # checked: absent from docs and from live envs). To settle each: run
+    # `env | grep _TRANSCRIPT_PATH` inside a hook and inside a shell-tool call.
+    ("gemini", ("GEMINI_TRANSCRIPT_PATH",)),
+    ("droid", ("FACTORY_TRANSCRIPT_PATH", "DROID_TRANSCRIPT_PATH")),
+    ("qoder", ("QODER_TRANSCRIPT_PATH",)),
+    ("codebuddy", ("CODEBUDDY_TRANSCRIPT_PATH",)),
+)
+_ENV_PLATFORM_ALIASES = {
+    "claude-code": "claude",
+    "factory": "droid",
+    "factory-ai": "droid",
+    "github-copilot": "copilot",
+}
+# ZCode intentionally reuses Claude's session env var name. Hooks know the host
+# is ZCode, while later shell commands see only the shared env name and resolve
+# it through the claude entry. Canonicalize both paths to one runtime filename.
+_CONTEXT_KEY_PLATFORM_ALIASES = {
+    "zcode": "claude",
+    # Factory Droid's config directory is `.factory/`, so a hook that names its
+    # platform after the directory it was installed in reports "factory". Its
+    # sibling hooks report "droid". One runtime filename either way.
+    "factory": "droid",
+}
+
+
+@dataclass(frozen=True)
+class ActiveTask:
+    """Resolved active task state."""
+
+    task_path: str | None
+    source_type: str
+    context_key: str | None = None
+    stale: bool = False
+
+    @property
+    def source(self) -> str:
+        """Human-readable source label."""
+        if self.source_type == "session" and self.context_key:
+            return f"session:{self.context_key}"
+        if self.source_type == "session-fallback" and self.context_key:
+            return f"session-fallback:{self.context_key}"
+        return self.source_type
+
+
+def normalize_task_ref(task_ref: str) -> str:
+    """Normalize a task ref for stable storage and comparison."""
+    normalized = task_ref.strip()
+    if not normalized:
+        return ""
+
+    path_obj = Path(normalized)
+    if path_obj.is_absolute():
+        return str(path_obj)
+
+    normalized = normalized.replace("\\", "/")
+    while normalized.startswith("./"):
+        normalized = normalized[2:]
+
+    if normalized.startswith(f"{DIR_TASKS}/"):
+        return f"{DIR_WORKFLOW}/{normalized}"
+
+    return normalized
+
+
+def resolve_task_ref(task_ref: str, repo_root: Path) -> Path | None:
+    """Resolve a task ref to an absolute task directory inside the repo.
+
+    Mirrors `paths.resolve_task_ref` (same containment check). Duplicated
+    rather than imported because this module is loaded standalone — hooks add
+    it to `sys.path` directly — so it stays zero-relative-import on purpose.
+    """
+    normalized = normalize_task_ref(task_ref)
+    if not normalized:
+        return None
+
+    path_obj = Path(normalized)
+    if path_obj.is_absolute():
+        candidate = path_obj
+    elif normalized.startswith(f"{DIR_WORKFLOW}/"):
+        candidate = repo_root / path_obj
+    else:
+        candidate = repo_root / DIR_WORKFLOW / DIR_TASKS / path_obj
+
+    # Both sides are resolved because repo_root itself may sit behind a
+    # symlink (/tmp on macOS does), and resolve() is what collapses `..`
+    # instead of leaving it for a lexical relative_to() to wave through.
+    try:
+        resolved = candidate.resolve()
+        root = repo_root.resolve()
+    except OSError:
+        return None
+
+    try:
+        resolved.relative_to(root)
+    except ValueError:
+        return None
+
+    return resolved
+
+
+def _runtime_sessions_dir(repo_root: Path) -> Path:
+    return repo_root / DIR_WORKFLOW / DIR_RUNTIME / DIR_SESSIONS
+
+
+def _sanitize_key(raw: str) -> str:
+    safe = re.sub(r"[^A-Za-z0-9._-]+", "_", raw.strip())
+    safe = safe.strip("._-")
+    return safe[:160] if safe else ""
+
+
+def _hash_value(raw: str) -> str:
+    return hashlib.sha256(raw.encode("utf-8")).hexdigest()[:24]
+
+
+def _as_dict(value: Any) -> dict[str, Any] | None:
+    return value if isinstance(value, dict) else None
+
+
+def _string_value(value: Any) -> str | None:
+    if isinstance(value, str):
+        stripped = value.strip()
+        return stripped or None
+    return None
+
+
+def _lookup_string(data: dict[str, Any], keys: tuple[str, ...]) -> str | None:
+    for key in keys:
+        value = _string_value(data.get(key))
+        if value:
+            return value
+
+    for nested_key in _NESTED_KEYS:
+        nested = _as_dict(data.get(nested_key))
+        if not nested:
+            continue
+        value = _lookup_string(nested, keys)
+        if value:
+            return value
+
+    return None
+
+
+def _detect_platform(platform_input: dict[str, Any] | None, platform: str | None) -> str:
+    if platform:
+        return _sanitize_key(platform) or "session"
+    if platform_input:
+        for key in ("_trellis_platform", "trellis_platform", "platform", "source"):
+            value = _string_value(platform_input.get(key))
+            if value:
+                return _sanitize_key(value) or "session"
+        if _string_value(platform_input.get("cursor_version")):
+            return "cursor"
+    return "session"
+
+
+def _context_key(platform_name: str, kind: str, value: str) -> str:
+    platform_name = _CONTEXT_KEY_PLATFORM_ALIASES.get(platform_name, platform_name)
+    if kind == "transcript":
+        return f"{platform_name}_transcript_{_hash_value(value)}"
+    safe_value = _sanitize_key(value)
+    if safe_value:
+        return f"{platform_name}_{safe_value}"
+    return f"{platform_name}_{_hash_value(value)}"
+
+
+def _iter_env_keys(
+    env_keys: tuple[tuple[str, tuple[str, ...]], ...],
+    platform_name: str | None,
+) -> tuple[tuple[str, tuple[str, ...]], ...]:
+    """Narrow an env-key table to one platform, or return all of it.
+
+    A platform with no entry yields an empty tuple, and the caller's `for` loop
+    simply does not run. That is the normal case, not an error: platforms with
+    no verified env var name are deliberately absent from these tables.
+    """
+    if not platform_name:
+        return env_keys
+    matched = tuple((name, keys) for name, keys in env_keys if name == platform_name)
+    return matched
+
+
+def _env_platform_name(platform_name: str | None) -> str | None:
+    if not platform_name or platform_name == "session":
+        return None
+    return _ENV_PLATFORM_ALIASES.get(platform_name, platform_name)
+
+
+def _lookup_env_context_key(platform_name: str | None) -> str | None:
+    """Resolve a context key from platform-provided environment variables.
+
+    Hooks pass `TRELLIS_CONTEXT_ID` to subprocesses they launch, but an AI-run
+    shell command can only see session identity if the host platform exports it
+    in the command environment. These names are best-effort adapters; if none
+    are present, there is no session-scoped active task.
+    """
+    env_platform_name = _env_platform_name(platform_name)
+
+    for name, keys in _iter_env_keys(_ENV_SESSION_KEYS, env_platform_name):
+        for key in keys:
+            value = _string_value(os.environ.get(key))
+            if value:
+                return _context_key(name, "session", value)
+
+    for name, keys in _iter_env_keys(_ENV_CONVERSATION_KEYS, env_platform_name):
+        for key in keys:
+            value = _string_value(os.environ.get(key))
+            if value:
+                return _context_key(name, "conversation", value)
+
+    for name, keys in _iter_env_keys(_ENV_TRANSCRIPT_KEYS, env_platform_name):
+        for key in keys:
+            value = _string_value(os.environ.get(key))
+            if value:
+                return _context_key(name, "transcript", value)
+
+    return None
+
+
+def _find_repo_root_from_cwd() -> Path | None:
+    current = Path.cwd().resolve()
+    while True:
+        if (current / DIR_WORKFLOW).is_dir():
+            return current
+        if current == current.parent:
+            return None
+        current = current.parent
+
+
+def _shell_ticket_dirs(repo_root: Path) -> tuple[Path, ...]:
+    runtime_dir = repo_root / DIR_WORKFLOW / DIR_RUNTIME
+    return (
+        runtime_dir / DIR_SHELL_TICKETS,
+        runtime_dir / DIR_LEGACY_CURSOR_SHELL_TICKETS,
+    )
+
+
+def _remove_file(path: Path) -> bool:
+    try:
+        path.unlink()
+        return True
+    except OSError:
+        return False
+
+
+def _task_refs_match(left: str | None, right: str | None, repo_root: Path) -> bool:
+    if not left or not right:
+        return False
+    left_path = resolve_task_ref(left, repo_root)
+    right_path = resolve_task_ref(right, repo_root)
+    if left_path is not None and right_path is not None:
+        return left_path == right_path
+    return normalize_task_ref(left) == normalize_task_ref(right)
+
+
+def _pending_ticket_matches_args(ticket: dict[str, Any], repo_root: Path) -> bool:
+    if Path(sys.argv[0]).name != "task.py":
+        return False
+    args = tuple(sys.argv[1:])
+    if not args:
+        return False
+
+    command_name = args[0]
+    if command_name not in TASK_SESSION_COMMANDS:
+        return False
+
+    subcommands = ticket.get("subcommands")
+    if not isinstance(subcommands, list):
+        return False
+
+    for subcommand in subcommands:
+        if not isinstance(subcommand, dict):
+            continue
+        if _string_value(subcommand.get("name")) != command_name:
+            continue
+        if command_name != "start":
+            return True
+        task_ref = args[1] if len(args) > 1 else None
+        if _task_refs_match(_string_value(subcommand.get("task_ref")), task_ref, repo_root):
+            return True
+
+    return False
+
+
+def _ticket_is_fresh(ticket: dict[str, Any], ticket_path: Path, now: float) -> bool:
+    expires_at = ticket.get("expires_at_epoch")
+    if isinstance(expires_at, (int, float)) and expires_at < now:
+        _remove_file(ticket_path)
+        return False
+
+    created_at = ticket.get("created_at_epoch")
+    if isinstance(created_at, (int, float)):
+        if now - created_at <= SHELL_TICKET_TTL_SECONDS:
+            return True
+        _remove_file(ticket_path)
+        return False
+    return True
+
+
+def _ticket_cwd_matches_repo(ticket: dict[str, Any], repo_root: Path) -> bool:
+    cwd = _string_value(ticket.get("cwd"))
+    if not cwd:
+        return True
+    try:
+        Path(cwd).resolve().relative_to(repo_root)
+    except ValueError:
+        return False
+    return True
+
+
+def _matching_ticket_context_key(
+    ticket_path: Path,
+    repo_root: Path,
+    now: float,
+) -> str | None:
+    """Accept a ticket on its merits, never on which platform wrote it.
+
+    The `platform` field a ticket carries is debugging metadata; gating on it
+    was what kept this bridge invisible to every platform but Cursor.
+    """
+    ticket = _read_json(ticket_path)
+    if ticket is None:
+        return None
+    if not _ticket_is_fresh(ticket, ticket_path, now):
+        return None
+    if not _ticket_cwd_matches_repo(ticket, repo_root):
+        return None
+    if not _pending_ticket_matches_args(ticket, repo_root):
+        return None
+    return _string_value(ticket.get("context_key"))
+
+
+def _lookup_shell_ticket_context_key() -> str | None:
+    """Resolve session identity from a short-lived shell ticket.
+
+    No researched platform exports its session id into a shell child, but every
+    hook-capable one hands that id to a hook. So the hook that runs just before
+    a shell command writes a ticket, and this reads it back. A ticket counts
+    only when it is fresh, was written for this repo, and matches the `task.py`
+    subcommand now running — and only when exactly one fresh context key
+    matches. Two concurrent windows therefore both degrade rather than one
+    inheriting the other's pointer.
+    """
+    repo_root = _find_repo_root_from_cwd()
+    if repo_root is None:
+        return None
+
+    now = time.time()
+    candidates: set[str] = set()
+    for ticket_dir in _shell_ticket_dirs(repo_root):
+        if not ticket_dir.is_dir():
+            continue
+        for ticket_path in ticket_dir.glob("*.json"):
+            context_key = _matching_ticket_context_key(ticket_path, repo_root, now)
+            if context_key:
+                candidates.add(context_key)
+
+    if len(candidates) == 1:
+        return next(iter(candidates))
+    return None
+
+
+def resolve_context_key(
+    platform_input: dict[str, Any] | None = None,
+    platform: str | None = None,
+    *,
+    allow_environment_context: bool = True,
+) -> str | None:
+    """Resolve a stable session/window context key, if one is available.
+
+    `TRELLIS_CONTEXT_ID` is an explicit context-key override used by CLI
+    scripts and subprocesses. It does not store the task itself.
+    """
+    if allow_environment_context:
+        override = _string_value(os.environ.get("TRELLIS_CONTEXT_ID"))
+        if override:
+            return _sanitize_key(override) or _hash_value(override)
+
+    data = _as_dict(platform_input)
+    platform_name = _detect_platform(data, platform) if data or platform else None
+
+    if data:
+        session_id = _lookup_string(data, _SESSION_KEYS)
+        if session_id:
+            return _context_key(platform_name or "session", "session", session_id)
+
+        conversation_id = _lookup_string(data, _CONVERSATION_KEYS)
+        if conversation_id:
+            return _context_key(platform_name or "session", "conversation", conversation_id)
+
+        transcript_path = _lookup_string(data, _TRANSCRIPT_KEYS)
+        if transcript_path:
+            return _context_key(platform_name or "session", "transcript", transcript_path)
+
+    if allow_environment_context:
+        env_context_key = _lookup_env_context_key(platform_name)
+        if env_context_key:
+            return env_context_key
+
+    # Last in the chain on purpose: a platform that genuinely exports identity
+    # into the shell outranks a ticket, and no platform name gates the lookup.
+    if allow_environment_context:
+        return _lookup_shell_ticket_context_key()
+    return None
+
+
+def _read_json(path: Path) -> dict[str, Any] | None:
+    try:
+        data = json.loads(path.read_text(encoding="utf-8"))
+    except (FileNotFoundError, json.JSONDecodeError, OSError):
+        return None
+    return data if isinstance(data, dict) else None
+
+
+def _write_json(path: Path, data: dict[str, Any]) -> bool:
+    try:
+        path.parent.mkdir(parents=True, exist_ok=True)
+        path.write_text(
+            json.dumps(data, indent=2, ensure_ascii=False) + "\n",
+            encoding="utf-8",
+        )
+        return True
+    except OSError:
+        return False
+
+
+def _canonical_task_ref(task_path: str, repo_root: Path) -> str | None:
+    normalized = normalize_task_ref(task_path)
+    if not normalized:
+        return None
+    full_path = resolve_task_ref(normalized, repo_root)
+    if full_path is None or not full_path.is_dir():
+        return None
+    try:
+        return full_path.relative_to(repo_root.resolve()).as_posix()
+    except ValueError:
+        # resolve_task_ref already refused everything outside the repo, so this
+        # is unreachable. Refuse rather than fall back to an absolute path —
+        # that fallback is how an out-of-repo ref used to reach the session
+        # pointer and get replayed on every later turn.
+        return None
+
+
+def _active_from_ref(
+    task_ref: str | None,
+    repo_root: Path,
+    source_type: str,
+    context_key: str | None = None,
+) -> ActiveTask | None:
+    if not task_ref:
+        return None
+    resolved = resolve_task_ref(task_ref, repo_root)
+    stale = resolved is None or not resolved.is_dir()
+    return ActiveTask(task_ref, source_type, context_key, stale)
+
+
+def _context_path(repo_root: Path, context_key: str) -> Path:
+    return _runtime_sessions_dir(repo_root) / f"{context_key}.json"
+
+
+def resolve_active_task(
+    repo_root: Path,
+    platform_input: dict[str, Any] | None = None,
+    platform: str | None = None,
+    *,
+    allow_single_session_fallback: bool = True,
+    allow_environment_context: bool = True,
+) -> ActiveTask:
+    """Resolve the active task from session runtime state only.
+
+    A stale session task is returned as stale. Missing context identity or a
+    missing/empty session context falls back to single-session inference: if
+    exactly one session file exists in the runtime, return its task with
+    source_type="session-fallback" — covers pull-based platform sub-agents
+    (copilot, gemini, qoder) that don't inherit the parent's session id. ≥2
+    files or 0 files yield ActiveTask(None) — refuses to guess across windows.
+    """
+    context_key = resolve_context_key(
+        platform_input,
+        platform,
+        allow_environment_context=allow_environment_context,
+    )
+    if context_key:
+        context = _read_json(_context_path(repo_root, context_key)) or {}
+        task_ref = _string_value(context.get("current_task"))
+        active = _active_from_ref(task_ref, repo_root, "session", context_key)
+        if active:
+            return active
+
+    if allow_single_session_fallback:
+        fallback = _resolve_single_session_fallback(repo_root)
+        if fallback is not None:
+            return fallback
+
+    return ActiveTask(None, "none", context_key)
+
+
+def _resolve_single_session_fallback(repo_root: Path) -> ActiveTask | None:
+    """Return the task pointed at by the sole session file, if exactly one exists.
+
+    Used when context-key resolution fails (typical for class-2 platform
+    sub-agents). Returns None if 0 or ≥2 session files are present — refuses
+    to pick across windows so 04-21's multi-session isolation contract holds.
+    """
+    sessions_dir = _runtime_sessions_dir(repo_root)
+    if not sessions_dir.is_dir():
+        return None
+
+    session_files = sorted(sessions_dir.glob("*.json"))
+    if len(session_files) != 1:
+        return None
+
+    session_file = session_files[0]
+    context = _read_json(session_file) or {}
+    task_ref = _string_value(context.get("current_task"))
+    if not task_ref:
+        return None
+
+    fallback_key = session_file.stem
+    return _active_from_ref(task_ref, repo_root, "session-fallback", fallback_key)
+
+
+def _utc_now() -> str:
+    return datetime.now(timezone.utc).replace(microsecond=0).isoformat().replace("+00:00", "Z")
+
+
+def _context_metadata(
+    platform_input: dict[str, Any] | None,
+    platform: str | None,
+    context_key: str | None = None,
+) -> dict[str, Any]:
+    data = _as_dict(platform_input) or {}
+    platform_name = _detect_platform(data, platform)
+    if platform_name == "session" and context_key:
+        prefix = context_key.split("_", 1)[0]
+        if prefix in _KNOWN_PLATFORMS:
+            platform_name = prefix
+    metadata: dict[str, Any] = {
+        "platform": platform_name,
+        "last_seen_at": _utc_now(),
+    }
+    for key in (*_SESSION_KEYS, *_CONVERSATION_KEYS, *_TRANSCRIPT_KEYS):
+        value = _lookup_string(data, (key,))
+        if value:
+            metadata[key] = value
+    return metadata
+
+
+def set_active_task(
+    task_path: str,
+    repo_root: Path,
+    platform_input: dict[str, Any] | None = None,
+    platform: str | None = None,
+) -> ActiveTask | None:
+    """Set the active task in session scope.
+
+    Returns None when no context key is available; callers should surface a
+    user-facing error that explains how to provide session identity.
+    """
+    canonical = _canonical_task_ref(task_path, repo_root)
+    if canonical is None:
+        return None
+
+    context_key = resolve_context_key(platform_input, platform)
+    if not context_key:
+        return None
+
+    context_path = _context_path(repo_root, context_key)
+    context = _read_json(context_path) or {}
+    context.update(_context_metadata(platform_input, platform, context_key))
+    context["current_task"] = canonical
+    context.setdefault("current_run", None)
+    if not _write_json(context_path, context):
+        return None
+    return ActiveTask(canonical, "session", context_key)
+
+
+def clear_active_task(
+    repo_root: Path,
+    platform_input: dict[str, Any] | None = None,
+    platform: str | None = None,
+) -> ActiveTask:
+    """Clear the active task by deleting its resolved session context file."""
+    context_key = resolve_context_key(platform_input, platform)
+    if not context_key:
+        return ActiveTask(None, "none")
+
+    previous = resolve_active_task(repo_root, platform_input, platform)
+    if not previous.task_path or not previous.context_key:
+        return previous
+
+    context_path = _context_path(repo_root, previous.context_key)
+    if context_path.is_file():
+        _remove_file(context_path)
+    return previous
+
+
+def clear_task_from_sessions(task_path: str, repo_root: Path) -> int:
+    """Delete all session runtime files that point at a task."""
+    target = _canonical_task_ref(task_path, repo_root) or normalize_task_ref(task_path)
+    if not target:
+        return 0
+
+    cleared = 0
+    sessions_dir = _runtime_sessions_dir(repo_root)
+    if not sessions_dir.is_dir():
+        return cleared
+
+    for session_path in sessions_dir.glob("*.json"):
+        context = _read_json(session_path) or {}
+        current = _string_value(context.get("current_task"))
+        if not current:
+            continue
+        current_ref = _canonical_task_ref(current, repo_root) or normalize_task_ref(current)
+        if current_ref != target:
+            continue
+        if session_path.is_file() and _remove_file(session_path):
+            cleared += 1
+
+    return cleared
+
+
+def get_current_task_source(
+    repo_root: Path,
+    platform_input: dict[str, Any] | None = None,
+    platform: str | None = None,
+) -> tuple[str, str | None, str | None]:
+    """Return (`source_type`, `context_key`, `task_path`) for compatibility."""
+    active = resolve_active_task(repo_root, platform_input, platform)
+    return active.source_type, active.context_key, active.task_path