Send a caller identity to flex-auth so policy.enabled can flip

flex-auth's flex-auth-ops-warden pin (FLEX-WP-0016) TokenReviews the caller and
binds resource.system: ops-warden to system:serviceaccount:ops-warden:ops-warden.
policy.py posted /v1/check with no Authorization header, so the pin logs
"caller authentication warning" and can only run callerAuth.mode: warn — which,
under ADHOC-2026-08-17-T01, is exactly what blocks policy.enabled: true.

- policy.caller_auth (none | file | env | command) + src/warden/caller_identity.py:
  token resolved per call, never cached, written, or logged (ADR-0002)
- both check_sign_policy and check_fetch_policy attach the bearer header; an
  unobtainable token fails closed rather than retrying anonymously
- scripts/check_policy_caller_identity.py: read-only gate, prints length and a
  truncated fingerprint only, distinguishes 401 (audience/binding) from 403
- example config: caller_auth block, and flex_auth_url corrected — it pointed at
  flex-auth.flex-auth.svc, a Service that does not exist
- WARDEN-WP-0031, PolicyGatedSigning caller-identity section and flip sequence

Default stays mode: none, so behaviour is unchanged until an operator opts in.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

SCOPE.md

@@ -50,8 +50,12 @@ net-kingdom / info-tech-canon is owner-driven (tracked via coordination messages
 
 **Policy gate** is shipped on the caller side (WP-0007) with production registry
 and smoke evidence (WP-0009 archived). flex-auth published the `ssh-certificate`
-policy package (FLEX-WP-0006). `policy.enabled` remains **false** in production
+policy package (FLEX-WP-0006) and now runs the in-cluster pin
-until flex-auth is deployed to a reachable URL (flex-auth FLEX-WP-0007).
+`flex-auth-ops-warden` (FLEX-WP-0007 finished; FLEX-WP-0016). `policy.enabled`
+remains **false**: that pin is in `callerAuth.mode: warn` because ops-warden sent
+no caller identity. WP-0031 ships the calling side (`policy.caller_auth` +
+`scripts/check_policy_caller_identity.py`); the flip waits on an operator picking
+a token source and flex-auth then enforcing.
 
 **ops-bridge cert_command pilot** is shipped to pilot-ready (WP-0016): a read-only
 readiness gate (`scripts/check_tunnel_cert_readiness.py`) plus an opt-in offline
@@ -273,6 +277,7 @@ for the rest.
 | WP | Status | Focus |
 | --- | --- | --- |
 | WP-0027 | `backlog` | Tamper-resistant credential governance + mass rotation/lockdown (Strand B follow-on to WP-0026) |
+| WP-0031 | `active` | Calling-side identity for flex-auth — code + gate done; flip sequence waits on operator + flex-auth enforce |
 | WP-0030 | `proposed` | Delegation register — record intended owner + blocker on every interim lane, `warden route gaps`, promotion gate |
 
 Remaining production distance is also in other repos' lanes (see Known gaps).
@@ -281,7 +286,7 @@ Remaining production distance is also in other repos' lanes (see Known gaps).
 
 | Gap | Owner | Notes |
 | --- | --- | --- |
-| flex-auth production runtime + registry deploy | flex-auth | **FLEX-WP-0007** — unblocks `policy.enabled: true` |
+| `callerAuth.mode: enforce` on the ops-warden pin | flex-auth | **FLEX-WP-0016 T03** — waits on WARDEN-WP-0031 T04, then unblocks `policy.enabled: true` |
 | ops-bridge `cert_command` on live tunnels | ops-bridge | Playbook + readiness gate shipped (WP-0016); pilot cutover handed off, awaiting ops-bridge |
 | Principals sync warden ↔ railiance-infra | ops-warden + infra | `scripts/check_principals_drift.py` — operator runs periodically |
 | NK-WP-0009 joint SSH tutorial | net-kingdom | Parallel coordination track |
@@ -347,7 +352,8 @@ Remaining production distance is also in other repos' lanes (see Known gaps).
 - **Production sign:** verified 2026-06-18 (`history/2026-06-17-openbao-production-verify.md`)
 - **Access routing:** WP-0010 + WP-0011 shipped (`warden route`, pointer catalog)
 - **Policy gate:** caller shipped (WP-0007); registry + smoke complete (WP-0009 archived).
-  `policy.enabled: false` until flex-auth reachable (`FLEX-WP-0007`)
+  `policy.enabled: false` — the `flex-auth-ops-warden` pin is in `warn` until
+  ops-warden sends a caller identity (WP-0031)
 - **Workload posture:** WP-0015 shipped (standard, descriptors, `warden policy`,
   conformance checker, dev doubles); canon landing owner-driven
 - **ops-bridge cert_command:** WP-0016 shipped to pilot-ready (readiness gate +

examples/warden.production.example.yaml

@@ -19,9 +19,29 @@ state_dir: ~/.local/state/warden
 # Registry: registry/flex-auth/production_registry_snapshot.json (build from inventory).
 # See wiki/PolicyGatedSigning.md (operator checklist) and wiki/playbooks/operator-openbao-token-hygiene.md
 policy:
+  # Keep false until the flex-auth ops-warden pin runs callerAuth.mode: enforce.
+  # Gate: python scripts/check_policy_caller_identity.py  (WARDEN-WP-0031)
   enabled: false
-  flex_auth_url: http://flex-auth.flex-auth.svc.cluster.local:8080
+  # The in-cluster pin for ops-warden's signing policy (FLEX-WP-0016). A bare
+  # flex-auth.flex-auth.svc Service does not exist. From a workstation, reach it
+  # through a port-forward or tunnel and point this at that local address.
+  flex_auth_url: http://flex-auth-ops-warden.flex-auth.svc.cluster.local:8080
   fail_closed: true
   tenant: tenant:platform
   subject_env: WARDEN_POLICY_SUBJECT
   system: ops-warden
+  # How ops-warden proves it is ops-warden. flex-auth TokenReviews this bearer
+  # token and requires the principal system:serviceaccount:ops-warden:ops-warden
+  # for resource.system: ops-warden. Mode none sends no header, which is what
+  # holds the pin in warn.
+  caller_auth:
+    mode: none  # none | file | env | command
+    # In-cluster PEP — projected ServiceAccount token, audience-bound:
+    # mode: file
+    # token_path: /var/run/secrets/flex-auth/token
+    # Workstation — mint a short-lived bound token per call:
+    # mode: command
+    # command: kubectl create token ops-warden -n ops-warden
+    #   --audience flex-auth --duration 10m
+    token_env: WARDEN_POLICY_CALLER_TOKEN
+    audience: flex-auth

scripts/check_policy_caller_identity.py

@@ -0,0 +1,189 @@
+#!/usr/bin/env python3
+"""Readiness gate for the flex-auth ``policy.enabled`` flip (WARDEN-WP-0031).
+
+flex-auth deployed ``flex-auth-ops-warden`` (FLEX-WP-0016) in ``callerAuth.mode:
+warn``: it authenticates the caller with a Kubernetes TokenReview and binds
+``resource.system: ops-warden`` to ``system:serviceaccount:ops-warden:ops-warden``,
+but a caller that sends no ``Authorization`` header only produces a
+``caller authentication warning`` and is still served. That pin cannot move to
+``enforce`` — and therefore ``policy.enabled: true`` cannot be set — until
+ops-warden's calling side actually presents a token.
+
+This script asserts the calling side *without* flipping anything:
+
+  * warden.yaml loads and ``policy.caller_auth.mode`` is not ``none``,
+  * a caller token can actually be obtained (file / env / command),
+  * (optional, ``--url``) a live ``/v1/check`` against the warn pin returns a
+    decision **and** the response is reached with the header attached.
+
+Exit 0 = ready to ask flex-auth to enforce, 1 = not ready, 2 = bad input.
+The token is never printed, logged, or written anywhere — only its length and a
+truncated SHA-256 fingerprint, which are safe to paste into a handoff message.
+
+Usage:
+  python scripts/check_policy_caller_identity.py [--config ~/.config/warden/warden.yaml]
+  python scripts/check_policy_caller_identity.py --url http://127.0.0.1:19090
+"""
+from __future__ import annotations
+
+import argparse
+import hashlib
+import sys
+from pathlib import Path
+from typing import List, Optional, Tuple
+
+_SRC = Path(__file__).resolve().parent.parent / "src"
+if _SRC.is_dir() and str(_SRC) not in sys.path:
+    sys.path.insert(0, str(_SRC))
+
+from warden.caller_identity import (  # noqa: E402
+    CallerIdentityError,
+    resolve_caller_token,
+)
+from warden.config import ConfigError, load_config  # noqa: E402
+
+Check = Tuple[str, str, str]
+
+
+def _fingerprint(token: str) -> str:
+    return "sha256:" + hashlib.sha256(token.encode()).hexdigest()[:12]
+
+
+def run_checks(config_path: Optional[Path], url: Optional[str]) -> List[Check]:
+    checks: List[Check] = []
+    try:
+        cfg = load_config(config_path)
+    except ConfigError as e:
+        return [("fail", "warden.yaml", str(e))]
+
+    policy = cfg.policy
+    checks.append(
+        ("ok", "warden.yaml", f"loaded; policy.enabled={str(policy.enabled).lower()}")
+    )
+
+    mode = policy.caller_auth.mode
+    if mode == "none":
+        checks.append(
+            (
+                "fail",
+                "caller_auth.mode",
+                "none — no Authorization header is sent; the flex-auth pin stays in warn",
+            )
+        )
+        return checks
+    checks.append(("ok", "caller_auth.mode", mode))
+
+    try:
+        token = resolve_caller_token(policy.caller_auth)
+    except CallerIdentityError as e:
+        checks.append(("fail", "caller token", str(e)))
+        return checks
+    assert token is not None
+    checks.append(
+        ("ok", "caller token", f"obtained, {len(token)} chars, {_fingerprint(token)}")
+    )
+
+    target = url or policy.flex_auth_url
+    if url is None and not policy.enabled:
+        checks.append(
+            (
+                "skip",
+                "live /v1/check",
+                f"policy.enabled=false; pass --url to smoke {target} anyway",
+            )
+        )
+        return checks
+
+    import httpx  # local import: the offline checks above must not need it
+
+    probe = {
+        "subject": {"id": "readiness", "type": "agt", "tenant": policy.tenant},
+        "action": "sign",
+        "resource": {
+            "id": "ssh-cert:actor/agt-state-hub-bridge",
+            "type": "ssh-certificate",
+            "system": policy.system,
+            "tenant": policy.tenant,
+        },
+        "context": {"readiness_probe": True},
+    }
+    try:
+        response = httpx.post(
+            target.rstrip("/") + "/v1/check",
+            json=probe,
+            headers={"Authorization": f"Bearer {token}"},
+            timeout=10.0,
+        )
+    except httpx.RequestError as e:
+        checks.append(("fail", "live /v1/check", f"unreachable at {target}: {e}"))
+        return checks
+
+    if response.status_code == 401:
+        checks.append(
+            (
+                "fail",
+                "live /v1/check",
+                "401 — the token was sent but flex-auth did not accept it "
+                "(check the TokenReview audience and the ServiceAccount binding)",
+            )
+        )
+    elif response.status_code == 403:
+        checks.append(
+            (
+                "fail",
+                "live /v1/check",
+                f"403 — authenticated, but the principal may not represent "
+                f"system {policy.system!r}",
+            )
+        )
+    elif response.status_code >= 400:
+        checks.append(
+            ("fail", "live /v1/check", f"HTTP {response.status_code} from {target}")
+        )
+    else:
+        try:
+            decision = response.json()
+        except ValueError:
+            checks.append(("fail", "live /v1/check", "non-JSON decision"))
+            return checks
+        effect = str(decision.get("effect", "?"))
+        decision_id = decision.get("id") or decision.get("request_id") or "?"
+        checks.append(
+            ("ok", "live /v1/check", f"HTTP 200, effect={effect}, decision={decision_id}")
+        )
+    return checks
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument("--config", type=Path, default=None, help="path to warden.yaml")
+    parser.add_argument(
+        "--url",
+        default=None,
+        help="flex-auth base URL to smoke (e.g. a port-forward of the warn pin)",
+    )
+    args = parser.parse_args()
+
+    checks = run_checks(args.config, args.url)
+    glyph = {"ok": "✓", "fail": "✗", "skip": "·"}
+    print("flex-auth caller-identity readiness\n")
+    for status, label, detail in checks:
+        print(f"  {glyph[status]} {label}: {detail}")
+
+    failed = [c for c in checks if c[0] == "fail"]
+    if failed:
+        print(
+            f"\nNOT READY — {len(failed)} check(s) failed. "
+            "Do not ask flex-auth to enforce, and do not set policy.enabled: true."
+        )
+        return 1
+    print(
+        "\nREADY — the calling side presents an identity. Next: tell flex-auth to set "
+        "callerAuth.mode: enforce on flex-auth-ops-warden, re-run this check, then set "
+        "policy.enabled: true with fail_closed: true."
+    )
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())

src/warden/caller_identity.py

@@ -0,0 +1,92 @@
+"""Caller identity for ops-warden's outbound flex-auth policy calls.
+
+flex-auth's `flex-auth-ops-warden` pin (FLEX-WP-0016) authenticates the *caller*
+before it evaluates the request: `Authorization: Bearer <token>` is passed to a
+Kubernetes TokenReview, and `resource.system: ops-warden` is bound to the
+principal `system:serviceaccount:ops-warden:ops-warden`. Until ops-warden sends
+that header, the pin logs `caller authentication warning` and can only run in
+`warn` mode — which is why `policy.enabled` cannot flip.
+
+This module resolves the token at call time and hands it straight to the request.
+Nothing is cached to disk, logged, or echoed: ops-warden carries the value, it
+does not hold it (ADR-0002).
+"""
+from __future__ import annotations
+
+import os
+import subprocess
+
+from warden.config import CallerAuthConfig
+
+
+class CallerIdentityError(Exception):
+    """Raised when a caller token was configured but could not be obtained."""
+
+
+def resolve_caller_token(cfg: CallerAuthConfig) -> str | None:
+    """Return the bearer token for flex-auth, or None when mode is ``none``.
+
+    Raises CallerIdentityError when a token was configured but is unavailable.
+    The token itself never appears in an exception message.
+    """
+    mode = cfg.mode
+    if mode == "none":
+        return None
+
+    if mode == "file":
+        if cfg.token_path is None:
+            raise CallerIdentityError("caller_auth mode 'file' has no token_path")
+        try:
+            token = cfg.token_path.read_text()
+        except OSError as e:
+            raise CallerIdentityError(
+                f"caller token file unreadable: {cfg.token_path} ({e.strerror})"
+            ) from e
+    elif mode == "env":
+        token = os.environ.get(cfg.token_env, "")
+        if not token.strip():
+            raise CallerIdentityError(
+                f"caller token env {cfg.token_env} is unset or empty"
+            )
+    elif mode == "command":
+        if not cfg.command:
+            raise CallerIdentityError("caller_auth mode 'command' has no command")
+        try:
+            result = subprocess.run(
+                cfg.command,
+                capture_output=True,
+                text=True,
+                timeout=30,
+                check=False,
+            )
+        except FileNotFoundError as e:
+            raise CallerIdentityError(
+                f"caller token command not found: {cfg.command[0]}"
+            ) from e
+        except subprocess.TimeoutExpired as e:
+            raise CallerIdentityError("caller token command timed out") from e
+        if result.returncode != 0:
+            stderr = (result.stderr or "").strip().splitlines()
+            detail = stderr[-1] if stderr else f"exit {result.returncode}"
+            raise CallerIdentityError(f"caller token command failed: {detail}")
+        token = result.stdout
+    else:
+        raise CallerIdentityError(f"unsupported caller_auth mode {mode!r}")
+
+    token = token.strip()
+    if not token:
+        raise CallerIdentityError(f"caller_auth mode {mode!r} produced an empty token")
+    if any(ch.isspace() for ch in token):
+        # flex-auth rejects a bearer token containing whitespace outright.
+        raise CallerIdentityError(
+            f"caller_auth mode {mode!r} produced a token containing whitespace"
+        )
+    return token
+
+
+def caller_auth_headers(cfg: CallerAuthConfig) -> dict[str, str]:
+    """Headers to attach to a flex-auth /v1/check call ({} when unauthenticated)."""
+    token = resolve_caller_token(cfg)
+    if token is None:
+        return {}
+    return {"Authorization": f"Bearer {token}"}

src/warden/config.py

@@ -2,9 +2,10 @@
 from __future__ import annotations
 
 import os
+import shlex
 from dataclasses import dataclass, field
 from pathlib import Path
-from typing import Dict, Optional
+from typing import Dict, List, Optional
 
 import yaml
 
@@ -13,6 +14,35 @@ class ConfigError(Exception):
     """Raised when config is invalid or missing."""
 
 
+@dataclass
+class CallerAuthConfig:
+    """How ops-warden proves *its own* identity to flex-auth (FLEX-WP-0016).
+
+    flex-auth's ops-warden pin authenticates the caller with a Kubernetes
+    TokenReview and binds ``resource.system: ops-warden`` to the principal
+    ``system:serviceaccount:ops-warden:ops-warden``. A workstation ``warden
+    sign`` is not a ServiceAccount, so the token has to come from somewhere:
+
+    ``none``     send no ``Authorization`` header (pre-FLEX-WP-0016 behaviour;
+                 accepted only while that pin runs ``callerAuth.mode: warn``)
+    ``file``     read a projected ServiceAccount token from ``token_path``
+                 (in-cluster PEP, audience-bound by the projection)
+    ``env``      read the token from ``token_env``
+    ``command``  run ``command`` and use its stdout, e.g.
+                 ``kubectl create token ops-warden -n ops-warden
+                 --audience flex-auth --duration 10m``
+
+    ops-warden never stores the token: it is read, sent, and dropped
+    (ADR-0002 — transparent conduit, not a broker).
+    """
+
+    mode: str = "none"
+    token_path: Optional[Path] = None
+    token_env: str = "WARDEN_POLICY_CALLER_TOKEN"
+    command: Optional[List[str]] = None
+    audience: str = "flex-auth"
+
+
 @dataclass
 class PolicyConfig:
     enabled: bool = False
@@ -21,6 +51,7 @@ class PolicyConfig:
     tenant: str = "tenant:platform"
     subject_env: str = "WARDEN_POLICY_SUBJECT"
     system: str = "ops-warden"
+    caller_auth: "CallerAuthConfig" = field(default_factory=lambda: CallerAuthConfig())
 
 
 @dataclass
@@ -117,6 +148,33 @@ def load_config(path: Optional[Path] = None) -> WardenConfig:
     )
 
     policy_raw = raw.get("policy") or {}
+    caller_raw = policy_raw.get("caller_auth") or {}
+    caller_command = caller_raw.get("command")
+    if isinstance(caller_command, str):
+        caller_command = shlex.split(caller_command)
+    elif caller_command is not None:
+        caller_command = [str(part) for part in caller_command]
+    caller_token_path = caller_raw.get("token_path")
+    caller_cfg = CallerAuthConfig(
+        mode=str(caller_raw.get("mode", "none")).strip().lower(),
+        token_path=(
+            Path(os.path.expanduser(str(caller_token_path)))
+            if caller_token_path
+            else None
+        ),
+        token_env=str(caller_raw.get("token_env", "WARDEN_POLICY_CALLER_TOKEN")),
+        command=caller_command,
+        audience=str(caller_raw.get("audience", "flex-auth")),
+    )
+    if caller_cfg.mode not in {"none", "file", "env", "command"}:
+        raise ConfigError(
+            f"policy.caller_auth.mode must be none|file|env|command, "
+            f"got {caller_cfg.mode!r}"
+        )
+    if caller_cfg.mode == "file" and caller_cfg.token_path is None:
+        raise ConfigError("policy.caller_auth.token_path is required for mode: file")
+    if caller_cfg.mode == "command" and not caller_cfg.command:
+        raise ConfigError("policy.caller_auth.command is required for mode: command")
     policy_cfg = PolicyConfig(
         enabled=bool(policy_raw.get("enabled", False)),
         flex_auth_url=str(policy_raw.get("flex_auth_url", "http://127.0.0.1:8080")),
@@ -124,6 +182,7 @@ def load_config(path: Optional[Path] = None) -> WardenConfig:
         tenant=str(policy_raw.get("tenant", "tenant:platform")),
         subject_env=str(policy_raw.get("subject_env", "WARDEN_POLICY_SUBJECT")),
         system=str(policy_raw.get("system", "ops-warden")),
+        caller_auth=caller_cfg,
     )
 
     return WardenConfig(

src/warden/policy.py

@@ -8,6 +8,7 @@ from pathlib import Path
 import httpx
 
 from warden.ca import CAError
+from warden.caller_identity import CallerIdentityError, caller_auth_headers
 from warden.config import PolicyConfig
 from warden.models import CertSpec
 
@@ -19,6 +20,21 @@ def pubkey_fingerprint(pubkey_path: Path) -> str:
     return f"sha256:{digest}"
 
 
+def _caller_headers(cfg: PolicyConfig) -> dict[str, str]:
+    """Bearer header identifying ops-warden itself to flex-auth (FLEX-WP-0016).
+
+    When the token cannot be obtained we refuse the call under ``fail_closed``
+    rather than silently falling back to an unauthenticated request — an
+    unauthenticated call is exactly what keeps the flex-auth pin in ``warn``.
+    """
+    try:
+        return caller_auth_headers(cfg.caller_auth)
+    except CallerIdentityError as e:
+        if cfg.fail_closed:
+            raise CAError(f"flex-auth caller identity unavailable: {e}") from e
+        return {}
+
+
 def _subject_id(cfg: PolicyConfig, spec: CertSpec) -> str:
     return os.environ.get(cfg.subject_env, "").strip() or spec.actor_name
 
@@ -60,8 +76,9 @@ def check_sign_policy(cfg: PolicyConfig, spec: CertSpec) -> str | None:
     }
 
     url = cfg.flex_auth_url.rstrip("/") + "/v1/check"
+    headers = _caller_headers(cfg)
     try:
-        response = httpx.post(url, json=request, timeout=10.0)
+        response = httpx.post(url, json=request, headers=headers, timeout=10.0)
         response.raise_for_status()
     except httpx.HTTPStatusError as e:
         if cfg.fail_closed:
@@ -120,8 +137,9 @@ def check_fetch_policy(
     }
 
     url = cfg.flex_auth_url.rstrip("/") + "/v1/check"
+    headers = _caller_headers(cfg)
     try:
-        response = httpx.post(url, json=request, timeout=10.0)
+        response = httpx.post(url, json=request, headers=headers, timeout=10.0)
         response.raise_for_status()
     except httpx.HTTPStatusError as e:
         if cfg.fail_closed:

tests/test_policy.py

@@ -1,4 +1,5 @@
 """Tests for warden.policy — flex-auth gate."""
+import subprocess
 from pathlib import Path
 from unittest.mock import MagicMock, patch
 
@@ -6,7 +7,12 @@ import httpx
 import pytest
 
 from warden.ca import CAError
-from warden.config import PolicyConfig
+from warden.caller_identity import (
+    CallerIdentityError,
+    caller_auth_headers,
+    resolve_caller_token,
+)
+from warden.config import CallerAuthConfig, PolicyConfig
 from warden.models import ActorType, CertSpec
 from warden.policy import check_sign_policy, pubkey_fingerprint
 
@@ -138,3 +144,122 @@ def test_subject_from_env(tmp_path, monkeypatch):
 
     body = post.call_args[1]["json"]
     assert body["subject"]["id"] == "iam:bernd"
+
+# --- caller identity (FLEX-WP-0016 / WARDEN-WP-0031) -----------------------
+
+def test_caller_auth_none_sends_no_header():
+    assert caller_auth_headers(CallerAuthConfig()) == {}
+
+
+def test_caller_auth_file_reads_projected_token(tmp_path):
+    token_file = tmp_path / "token"
+    token_file.write_text("sa-token-value\n")
+    cfg = CallerAuthConfig(mode="file", token_path=token_file)
+    assert caller_auth_headers(cfg) == {"Authorization": "Bearer sa-token-value"}
+
+
+def test_caller_auth_file_missing_raises(tmp_path):
+    cfg = CallerAuthConfig(mode="file", token_path=tmp_path / "absent")
+    with pytest.raises(CallerIdentityError, match="unreadable"):
+        resolve_caller_token(cfg)
+
+
+def test_caller_auth_env_mode(monkeypatch):
+    monkeypatch.setenv("WARDEN_POLICY_CALLER_TOKEN", " env-token ")
+    assert resolve_caller_token(CallerAuthConfig(mode="env")) == "env-token"
+    monkeypatch.setenv("WARDEN_POLICY_CALLER_TOKEN", "")
+    with pytest.raises(CallerIdentityError, match="unset or empty"):
+        resolve_caller_token(CallerAuthConfig(mode="env"))
+
+
+def test_caller_auth_command_mode_uses_stdout(monkeypatch):
+    cfg = CallerAuthConfig(mode="command", command=["kubectl", "create", "token"])
+
+    def fake_run(cmd, **kwargs):
+        assert cmd == cfg.command
+        return subprocess.CompletedProcess(cmd, 0, stdout="minted-token\n", stderr="")
+
+    monkeypatch.setattr(subprocess, "run", fake_run)
+    assert resolve_caller_token(cfg) == "minted-token"
+
+
+def test_caller_auth_command_failure_message_excludes_token(monkeypatch):
+    cfg = CallerAuthConfig(mode="command", command=["kubectl", "create", "token"])
+
+    def fake_run(cmd, **kwargs):
+        return subprocess.CompletedProcess(cmd, 1, stdout="", stderr="error: forbidden\n")
+
+    monkeypatch.setattr(subprocess, "run", fake_run)
+    with pytest.raises(CallerIdentityError, match="error: forbidden"):
+        resolve_caller_token(cfg)
+
+
+def test_caller_auth_rejects_whitespace_token(tmp_path):
+    token_file = tmp_path / "token"
+    token_file.write_text("two words")
+    cfg = CallerAuthConfig(mode="file", token_path=token_file)
+    with pytest.raises(CallerIdentityError, match="whitespace"):
+        resolve_caller_token(cfg)
+
+
+def test_sign_policy_sends_authorization_header(tmp_path, monkeypatch):
+    """The header flex-auth's ops-warden pin needs to leave warn mode."""
+    from warden import policy as policy_mod
+
+    token_file = tmp_path / "token"
+    token_file.write_text("sa-token-value")
+    pubkey = tmp_path / "id.pub"
+    pubkey.write_text("ssh-ed25519 AAAA test\n")
+
+    cfg = PolicyConfig(
+        enabled=True,
+        caller_auth=CallerAuthConfig(mode="file", token_path=token_file),
+    )
+    spec = CertSpec(
+        actor_name="agt-state-hub-bridge",
+        actor_type=ActorType.AGT,
+        principals=["agt"],
+        ttl_hours=24,
+        pubkey_path=pubkey,
+    )
+
+    seen = {}
+
+    class _Response:
+        status_code = 200
+
+        def raise_for_status(self):
+            return None
+
+        def json(self):
+            return {"effect": "allow", "id": "decision:49350f1064f674d7"}
+
+    def fake_post(url, json=None, headers=None, timeout=None):
+        seen["headers"] = headers
+        return _Response()
+
+    monkeypatch.setattr(policy_mod.httpx, "post", fake_post)
+    assert policy_mod.check_sign_policy(cfg, spec) == "decision:49350f1064f674d7"
+    assert seen["headers"] == {"Authorization": "Bearer sa-token-value"}
+
+
+def test_sign_policy_fail_closed_when_caller_token_unavailable(tmp_path):
+    from warden.ca import CAError
+    from warden import policy as policy_mod
+
+    pubkey = tmp_path / "id.pub"
+    pubkey.write_text("ssh-ed25519 AAAA test\n")
+    cfg = PolicyConfig(
+        enabled=True,
+        fail_closed=True,
+        caller_auth=CallerAuthConfig(mode="file", token_path=tmp_path / "absent"),
+    )
+    spec = CertSpec(
+        actor_name="agt-state-hub-bridge",
+        actor_type=ActorType.AGT,
+        principals=["agt"],
+        ttl_hours=24,
+        pubkey_path=pubkey,
+    )
+    with pytest.raises(CAError, match="caller identity unavailable"):
+        policy_mod.check_sign_policy(cfg, spec)

wiki/PolicyGatedSigning.md

@@ -66,6 +66,8 @@ policy:
   tenant: tenant:platform
   subject_env: WARDEN_POLICY_SUBJECT
   system: ops-warden
+  caller_auth:            # how ops-warden identifies itself — see "Caller identity"
+    mode: none
 ```
 
 | Key | Default | Description |
@@ -76,6 +78,7 @@ policy:
 | `tenant` | `tenant:platform` | Tenant sent in subject and resource |
 | `subject_env` | `WARDEN_POLICY_SUBJECT` | Env var for IAM subject id override |
 | `system` | `ops-warden` | Resource system identifier |
+| `caller_auth.mode` | `none` | `none` \| `file` \| `env` \| `command` — source of the `Authorization` bearer token flex-auth TokenReviews ([Caller identity](#caller-identity-warden-wp-0031)) |
 
 Set `WARDEN_POLICY_SUBJECT` to the caller's IAM profile `sub` when available.
 If unset, the actor name is used as subject id.
@@ -196,6 +199,75 @@ Evidence: `history/2026-06-23-flex-auth-policy-gate-production-smoke.md`.
 
 ---
 
+## Caller identity (WARDEN-WP-0031)
+
+flex-auth authenticates the **caller** before it evaluates the request. The
+in-cluster pin `flex-auth-ops-warden` (FLEX-WP-0016) passes the bearer token to a
+Kubernetes TokenReview and requires the principal
+`system:serviceaccount:ops-warden:ops-warden` for any request carrying
+`resource.system: ops-warden`.
+
+Until ops-warden sends that header the pin logs
+
+```
+caller authentication warning: caller is not authenticated
+```
+
+and can only run `callerAuth.mode: warn`. Per ADHOC-2026-08-17-T01,
+`policy.enabled` must not flip while `/v1/check` still answers unauthenticated
+callers — so **the missing header is what blocks the flip**, and it is
+ops-warden's to fix, not flex-auth's.
+
+### Configure a token source
+
+```yaml
+policy:
+  caller_auth:
+    mode: none          # none | file | env | command
+    token_path: /var/run/secrets/flex-auth/token   # mode: file
+    token_env: WARDEN_POLICY_CALLER_TOKEN          # mode: env
+    command: kubectl create token ops-warden -n ops-warden --audience flex-auth --duration 10m
+    audience: flex-auth
+```
+
+| Mode | Use it when |
+| --- | --- |
+| `none` | Default. No header — pre-FLEX-WP-0016 behaviour, only viable while the pin is in `warn` |
+| `file` | In-cluster PEP with a projected, audience-bound ServiceAccount token |
+| `env` | The token is already in the environment (CI, a wrapper) |
+| `command` | Workstation `warden sign` — mint a short-lived bound token per call |
+
+A workstation is not a ServiceAccount, which is why `command` exists. The token is
+read, sent, and dropped: never cached, written, or logged (ADR-0002). If a token
+is configured but cannot be obtained, the sign is **refused** under
+`fail_closed` — falling back to an anonymous call would defeat the gate.
+
+### Readiness gate
+
+```bash
+python scripts/check_policy_caller_identity.py                     # offline
+python scripts/check_policy_caller_identity.py --url http://127.0.0.1:19090   # port-forward of the warn pin
+```
+
+Exit 0 ready / 1 not ready / 2 bad input. It prints the token's length and a
+truncated fingerprint, never the value, so its output is safe to paste into a
+handoff message. A live `401` means the token was sent but rejected (audience or
+binding); `403` means it authenticated but may not represent `system: ops-warden`.
+
+### Flip sequence
+
+1. Configure `caller_auth`; `check_policy_caller_identity.py` exits 0 offline.
+2. Smoke against a port-forward of the warn pin. The evidence is the **absence**
+   of `caller authentication warning` in its log, not the `allow` — warn serves
+   unauthenticated callers too, so an allow alone proves nothing.
+3. Ask flex-auth to set `callerAuth.mode: enforce` (their FLEX-WP-0016 T03).
+4. Re-run the gate against the enforcing pin.
+5. Only then: `policy.enabled: true`, `fail_closed: true`.
+
+Reversing 3 and 5 401s every `warden sign`.
+
+---
+
 ## Production rollout
 
 **Keep `policy.enabled: false` until flex-auth is reachable** at `policy.flex_auth_url`
@@ -210,7 +282,9 @@ with `fail_closed: true`, unreachable flex-auth blocks all signs.
 | 3 | ops-warden | Regenerate registry from inventory: `scripts/build_flex_auth_registry.py` |
 | 4 | ops-warden | Local smoke: `./scripts/policy_gate_production_smoke.sh` |
 | 5 | operator | Vault smoke: `make credential-exec-ops-warden-smoke` in `railiance-platform` (or manual `SMOKE_VAULT=1` fallback) |
-| 6 | operator | Set `policy.flex_auth_url` in `~/.config/warden/warden.yaml` |
+| 6 | operator | Set `policy.flex_auth_url` in `~/.config/warden/warden.yaml` — the pin is `flex-auth-ops-warden.flex-auth.svc.cluster.local:8080`, reached from a workstation via port-forward or tunnel |
+| 6a | ops-warden | Configure `policy.caller_auth`; `scripts/check_policy_caller_identity.py` exits 0 (see **Caller identity**) |
+| 6b | flex-auth | Set `callerAuth.mode: enforce` on the ops-warden pin (FLEX-WP-0016 T03) |
 | 7 | operator | Set `policy.enabled: true`; keep `fail_closed: true` |
 | 8 | operator | Allow smoke: `warden sign <actor>` — `signatures.log` has `policy_decision_id` |
 | 9 | operator | Deny smoke: e.g. `--ttl` above max — CLI shows flex-auth `reason`, no cert |

workplans/WARDEN-WP-0031-policy-caller-identity.md

@@ -0,0 +1,133 @@
+---
+id: WARDEN-WP-0031
+type: workplan
+title: "Calling-side identity for flex-auth, so policy.enabled can flip"
+domain: infotech
+repo: ops-warden
+status: active
+owner: ops-warden
+topic_slug: netkingdom
+planning_priority: P1
+depends_on_workplans:
+  - WARDEN-WP-0007
+related_workplans:
+  - WARDEN-WP-0009
+created: "2026-08-19"
+updated: "2026-08-19"
+---
+
+# WARDEN-WP-0031 — Calling-side identity for flex-auth
+
+flex-auth shipped `flex-auth-ops-warden` (FLEX-WP-0016 T01/T02): an
+independently rollable in-cluster pin carrying ops-warden's production registry
+and policy package, on digest `sha256:138aa347…`, at
+`flex-auth-ops-warden.flex-auth.svc.cluster.local:8080`.
+
+It runs `callerAuth.mode: warn`, and it says why in its own logs:
+
+```
+caller authentication warning: caller is not authenticated
+```
+
+`src/warden/policy.py` posted `/v1/check` with **no `Authorization` header**.
+flex-auth authenticates the caller with a Kubernetes TokenReview and binds
+`resource.system: ops-warden` to `system:serviceaccount:ops-warden:ops-warden`;
+an unauthenticated caller can only be served in `warn`. So the pin cannot
+enforce, and per ADHOC-2026-08-17-T01 — `policy.enabled` must not flip anywhere
+while `/v1/check` still answers unauthenticated callers — `policy.enabled` stays
+false. The gap is ours, not flex-auth's, and this workplan closes it.
+
+Warn is also not A2 evidence: a request that succeeds because failures are
+downgraded proves nothing about the enforcing path.
+
+## Ownership
+
+| Concern | Owner |
+| --- | --- |
+| The pin, its digest, `callerAuth.mode` | flex-auth |
+| Sending a caller identity on `/v1/check` | **ops-warden** (this workplan) |
+| Choosing the token source on a given host | ops-warden operator |
+| `policy.enabled: true` in `warden.yaml` | ops-warden operator |
+
+## Design note — fail closed on identity too
+
+When a caller token is configured but cannot be obtained, `check_sign_policy`
+raises under `fail_closed` rather than retrying unauthenticated. Falling back to
+an anonymous call is precisely the behaviour that keeps the pin in `warn`; a
+gate that silently degrades to the ungated path is not a gate (ADR-0004's choke
+point argument, applied to ops-warden as a caller).
+
+## Tasks
+
+```task
+id: WARDEN-WP-0031-T01
+status: done
+priority: high
+```
+
+**Caller identity on the outbound policy call.** `policy.caller_auth` in
+`warden.yaml` (`mode: none | file | env | command`, `token_path`, `token_env`,
+`command`, `audience`); `src/warden/caller_identity.py` resolves the token at
+call time and never caches, logs, or echoes it (ADR-0002); both
+`check_sign_policy` and `check_fetch_policy` attach `Authorization: Bearer …`.
+Whitespace-bearing and empty tokens are rejected before the call, because
+flex-auth rejects them outright.
+
+Done 2026-08-19. `mode: none` remains the default, so behaviour is unchanged
+until an operator opts in. Tests in `tests/test_policy.py`.
+
+```task
+id: WARDEN-WP-0031-T02
+status: done
+priority: high
+```
+
+**Readiness gate.** `scripts/check_policy_caller_identity.py` — read-only:
+config loads, mode is not `none`, a token is actually obtainable, and with
+`--url` a live `/v1/check` against a port-forward of the warn pin. Prints the
+token's length and a truncated SHA-256 fingerprint only, so its output is safe
+to paste into a handoff. Exit 0 ready / 1 not ready / 2 bad input. Distinguishes
+401 (token not accepted — audience or binding) from 403 (authenticated but not
+allowed to represent `system: ops-warden`).
+
+Done 2026-08-19.
+
+```task
+id: WARDEN-WP-0031-T03
+status: done
+priority: medium
+```
+
+**Docs.** `examples/warden.production.example.yaml` gains the `caller_auth`
+block with both realistic sources, and its `flex_auth_url` is corrected — the
+example pointed at `flex-auth.flex-auth.svc.cluster.local`, a Service that does
+not exist. `wiki/PolicyGatedSigning.md` gains the caller-identity section and
+the flip sequence.
+
+Done 2026-08-19.
+
+```task
+id: WARDEN-WP-0031-T04
+status: wait
+priority: high
+```
+
+**Pick the token source and prove it against the warn pin.** Operator work on
+the real host: `kubectl create token` (workstation) or a projected token
+(in-cluster PEP), then
+`python scripts/check_policy_caller_identity.py --url http://127.0.0.1:19090`
+against a port-forward. Expect `effect=allow` for `agt-state-hub-bridge` while
+the warn log stops printing `caller authentication warning` — the absence of
+that line, not the allow, is the evidence.
+
+```task
+id: WARDEN-WP-0031-T05
+status: wait
+priority: high
+```
+
+**Sequence the flip.** Only after T04: tell flex-auth to set
+`callerAuth.mode: enforce` on `flex-auth-ops-warden` (their FLEX-WP-0016 T03),
+re-run the gate against the enforcing pin, and only then set
+`policy.enabled: true` with `fail_closed: true`. Flipping before enforce buys
+nothing; flipping before T04 401s every `warden sign`.