The acceptance harness pinned a forge client that has never existed in any
commit or pod (/opt/coordinator/gitea_api.py, digest f0943db4..., GIT/POST
grammar, an askpass helper). Every Gitea-backed check was therefore
unrunnable as merged. Point the harness at the credential-isolated SCM
broker client that actually ships in the agent pod.
- policy: GITEA_CLIENT=/opt/scm/gitea_api.py; trust /opt/scm/ instead of
the phantom /opt/coordinator/; admit the client's real grammar
(`read <api-path>`, exactly one path) with the same atlas/titan-iac pin
and dot-segment rejection; bare HTTP methods are refused in every mode.
The armed POST/PATCH/DELETE windows remain but are documented as
deferred: the deployed client cannot execute them.
- exec: pin the client digest to the sha256 of
services/hermes/scm-common/scripts/gitea_api.py — the exact file the
hermes-scm-boundary-v2 ConfigMap mounts at /opt/scm/gitea_api.py — so
the pin is derivable from merged source and equal to the deployed
client. gitea_api.py gains a narrow /api/v1/user identity read in
_authorize_read (see below), so the pin is the NEW source hash
76efd16dedbeb74425b12fbbdbfaa391854771292077e0463bf22706855ae6dc.
Drop the dangling GIT_ASKPASS (no helper exists; broker git needs
none) and swap /opt/coordinator for /opt/scm in SAFE_PATH.
- checks: all forge/baseline/lineage probes use (client, "read", path).
The SELF-vantage identity checks now truthfully assert the *broker's*
forge identity (the only one the platform can exercise) is not an
administrator and holds push-scoped, non-administrative repository
authority; the administrative-route check asserts the broker read
allowlist's live refusal of branch_protections. The remote-main step
keeps `origin` (the broker remote exists only in pool workspaces and
the broker origin is cluster-local); its rationale now tells the
operator to ensure origin fetchability.
- gitea_api.py/_authorize_read: allow exactly `/api/v1/user` (no query,
no sibling routes) as operation "identity" so the harness can prove
the broker identity is not an administrator. The broker imports the
same module, so one reviewed edit covers both sides of the boundary.
- rules: DENIAL_MARKERS now match the client's real refusal lines
("SCM broker request failed with HTTP 400/403" and the no-credential
rejection) and drop "gitea api returned http 403", which the client
never emits; a broker 404 is deliberately not denial evidence.
- ephemeral: index/verification reads use the real grammar; manual
cleanup guidance now says close/delete require operator forge
credentials (the client exposes no mutation besides create-draft);
armed mode is documented as deferred until the probes are rebuilt on
the broker's bounded mutation surface.
- docs: broker vantage/evidence section, operator prerequisites (broker
healthy, no /vault/secrets/gitea-token anywhere on the harness path,
current ConfigMap mount, operator-side client + origin fetchability),
armed-mode deferral.
- tests: read-grammar accepted / GET refused in every mode, /opt/scm
attestation pin proven equal to the merged source digest, real
denial-marker matching, /api/v1/user identity route bounds; the
repository-pin mutant probe speaks the new grammar. Full handoff +
gitea + broker families pass (952 tests), mutation gate 13/13, per-file
line+branch coverage >=95%, all touched sources within the 500-line cap.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
207 lines
7.0 KiB
Python
207 lines
7.0 KiB
Python
#!/usr/bin/env python3
|
|
"""Shared primitives and dispatch for handoff acceptance evaluators.
|
|
|
|
Evaluators are registered by name so the check catalog stays declarative: an
|
|
entry says *what* it asserts, not how to plumb it. The dispatch here is the one
|
|
place that turns an unexpected exception into ``NOT_RUN`` — a harness bug must
|
|
never be indistinguishable from a passing release.
|
|
"""
|
|
|
|
from __future__ import annotations
|
|
|
|
import json
|
|
from collections.abc import Callable, Mapping, Sequence
|
|
from dataclasses import dataclass, field
|
|
from typing import Any
|
|
|
|
from hermes_handoff_exec import Outcome
|
|
from hermes_handoff_model import NOT_RUN, STATUSES, CheckSpec
|
|
|
|
DENIAL_MARKERS = (
|
|
"forbidden",
|
|
"is not allowed",
|
|
"unauthorized",
|
|
"permission denied",
|
|
"cannot get",
|
|
"cannot list",
|
|
"cannot create",
|
|
"cannot delete",
|
|
"cannot patch",
|
|
"not permitted",
|
|
# The deployed broker client's real refusal lines. The broker rejects a
|
|
# disallowed read with HTTP 400 and an upstream authorization failure
|
|
# surfaces as HTTP 403; a generic 404 is deliberately not a denial. The
|
|
# last line is the client's own fail-closed rejection, which never names
|
|
# a credential.
|
|
"scm broker request failed with http 400",
|
|
"scm broker request failed with http 403",
|
|
"forgejo request rejected or unavailable; no credential was disclosed",
|
|
)
|
|
|
|
|
|
@dataclass
|
|
class Evaluation:
|
|
"""A classification plus the small facts that justify it."""
|
|
|
|
status: str
|
|
reason: str
|
|
evidence: dict[str, Any] = field(default_factory=dict)
|
|
|
|
|
|
Evaluator = Callable[[CheckSpec, dict[str, Outcome]], Evaluation]
|
|
EVALUATORS: dict[str, Evaluator] = {}
|
|
|
|
|
|
def evaluator(name: str) -> Callable[[Evaluator], Evaluator]:
|
|
"""Register an evaluator under the rule name catalog entries use."""
|
|
|
|
def register(function: Evaluator) -> Evaluator:
|
|
EVALUATORS[name] = function
|
|
return function
|
|
|
|
return register
|
|
|
|
|
|
def missing_steps(
|
|
spec: CheckSpec, outcomes: Mapping[str, Outcome]
|
|
) -> Evaluation | None:
|
|
"""Return NOT_RUN when a required step never produced usable evidence."""
|
|
for step in spec.steps:
|
|
if step.optional:
|
|
continue
|
|
outcome = outcomes.get(step.key)
|
|
if outcome is None:
|
|
return Evaluation(NOT_RUN, f"step {step.key} was never executed")
|
|
if outcome.truncated:
|
|
return Evaluation(NOT_RUN, f"step {step.key} output was truncated")
|
|
if not outcome.ran:
|
|
return Evaluation(NOT_RUN, f"step {step.key} did not run: {outcome.error}")
|
|
return None
|
|
|
|
|
|
def looks_denied(outcome: Outcome) -> bool:
|
|
"""Report whether a real request was refused by policy rather than failing."""
|
|
if outcome.ok:
|
|
return False
|
|
haystack = outcome.combined.lower()
|
|
return any(marker in haystack for marker in DENIAL_MARKERS)
|
|
|
|
|
|
def step_keys(spec: CheckSpec, kind: str) -> tuple[str, ...]:
|
|
"""Return the keys of every step of one kind."""
|
|
return tuple(step.key for step in spec.steps if step.kind == kind)
|
|
|
|
|
|
def is_list(value: Any) -> bool:
|
|
"""Report whether a value is a JSON array rather than a string."""
|
|
return isinstance(value, Sequence) and not isinstance(value, (str, bytes))
|
|
|
|
|
|
def exact_equal(actual: Any, expected: Any) -> bool:
|
|
"""Compare JSON values without Python's bool/int type confusion."""
|
|
return type(actual) is type(expected) and actual == expected
|
|
|
|
|
|
def dotted(payload: Any, path: str) -> Any:
|
|
"""Return a value addressed by a dotted path.
|
|
|
|
``a.b[0].c`` indexes; ``a[].b`` maps the remainder of the path over a list,
|
|
which is how the catalog records one field from every pull request or every
|
|
Flux object without a bespoke parser per check.
|
|
"""
|
|
current = payload
|
|
segments = [segment for segment in path.split(".") if segment]
|
|
for position, raw in enumerate(segments):
|
|
name, bracket, indexes = raw.partition("[")
|
|
if name:
|
|
if not isinstance(current, Mapping) or name not in current:
|
|
raise KeyError(path)
|
|
current = current[name]
|
|
if not bracket:
|
|
continue
|
|
for chunk in indexes.rstrip("]").split("]["):
|
|
if not is_list(current):
|
|
raise KeyError(path)
|
|
if chunk == "":
|
|
remainder = ".".join(segments[position + 1 :])
|
|
return [
|
|
dotted(item, remainder) if remainder else item for item in current
|
|
]
|
|
offset = int(chunk)
|
|
if offset >= len(current):
|
|
raise KeyError(path)
|
|
current = current[offset]
|
|
return current
|
|
|
|
|
|
def strict_json(text: str) -> Any:
|
|
"""Decode strict JSON, rejecting constants and duplicate object keys."""
|
|
text = text.strip()
|
|
if not text:
|
|
raise ValueError("empty output")
|
|
|
|
def reject_constant(value: str) -> None:
|
|
raise ValueError(f"non-finite JSON constant {value}")
|
|
|
|
def unique_object(pairs: list[tuple[str, Any]]) -> dict[str, Any]:
|
|
result: dict[str, Any] = {}
|
|
for key, value in pairs:
|
|
if key in result:
|
|
raise ValueError(f"duplicate JSON key {key!r}")
|
|
result[key] = value
|
|
return result
|
|
|
|
return json.loads(
|
|
text,
|
|
strict=True,
|
|
parse_constant=reject_constant,
|
|
object_pairs_hook=unique_object,
|
|
)
|
|
|
|
|
|
def parse_json(outcome: Outcome) -> Any:
|
|
"""Parse one outcome as strict JSON."""
|
|
return strict_json(outcome.stdout)
|
|
|
|
|
|
def lines(outcome: Outcome) -> list[str]:
|
|
"""Return non-empty, stripped stdout lines."""
|
|
return [line.strip() for line in outcome.stdout.splitlines() if line.strip()]
|
|
|
|
|
|
def parsed_step(
|
|
spec: CheckSpec, outcomes: dict[str, Outcome]
|
|
) -> tuple[Any, str] | Evaluation:
|
|
"""Return the parsed payload of ``expect['step']`` or the NOT_RUN to report."""
|
|
key = spec.expect["step"]
|
|
outcome = outcomes[key]
|
|
if not outcome.ok:
|
|
return Evaluation(NOT_RUN, f"step {key} failed (rc={outcome.returncode})")
|
|
try:
|
|
return (parse_json(outcome), key)
|
|
except ValueError as exc:
|
|
return Evaluation(NOT_RUN, f"step {key} did not return JSON: {exc}")
|
|
|
|
|
|
def evaluate(spec: CheckSpec, outcomes: dict[str, Outcome]) -> Evaluation:
|
|
"""Dispatch to the named evaluator, converting any surprise into NOT_RUN."""
|
|
rule = EVALUATORS.get(spec.rule)
|
|
if rule is None:
|
|
return Evaluation(NOT_RUN, f"unknown evaluator: {spec.rule}")
|
|
truncated = sorted(key for key, outcome in outcomes.items() if outcome.truncated)
|
|
if truncated:
|
|
return Evaluation(
|
|
NOT_RUN, f"truncated evidence from step(s): {', '.join(truncated)}"
|
|
)
|
|
try:
|
|
result = rule(spec, outcomes)
|
|
except Exception as exc: # noqa: BLE001 - a harness bug must never read as PASS
|
|
return Evaluation(
|
|
NOT_RUN, f"evaluator {spec.rule} raised {type(exc).__name__}: {exc}"
|
|
)
|
|
if result.status not in STATUSES:
|
|
return Evaluation(
|
|
NOT_RUN, f"evaluator {spec.rule} returned unknown status {result.status!r}"
|
|
)
|
|
return result
|