feat(P23): tagging standard + Wiz adapter + Kyverno adapter

Phase 23 (v1.7) — tagging standards and security adapters.

* schemas/tagging-standard.json (D-054): canonical required-tags schema
  (acdl:owner, acdl:contract, acdl:environment, acdl:cost-center).
* adapters/terraform/policy/custom_rules/acdl_tagging.py: Checkov custom
  rule (ACDL_TAG_NAMING) loaded via --external-checks-dir; closes D-043
  (synthetic SKIPPED record replaced by real PASS/FAIL records).
* checkov_adapter.py: removed _emit_tag_naming_skipped(), added
  ACDL_TAG_NAMING to RULE_MAP, updated docstring.
* scripts/run_platform.sh: both Checkov invocations pass
  --external-checks-dir adapters/terraform/policy/custom_rules/.
* adapters/wiz/ (D-052): Wiz adapter translating issue records to
  PolicyCheckResult (engine: "wiz"); graceful degradation emits
  WIZ_NOT_CONFIGURED SKIPPED when unconfigured; is_configured() gate.
* adapters/kyverno/ (D-053): Kyverno adapter translating PolicyReport
  results to PolicyCheckResult (engine: "kyverno"); ready but inactive
  for Terraform-only stacks; 3 sample ClusterPolicies in policies/.
* schemas/policy_check_result.schema.json: engine enum += "wiz".
* tests: fixtures + test_wiz_adapter.py (8 tests) + test_kyverno_adapter.py
  (13 tests); updated test_checkov_adapter.py to not expect the removed
  synthetic ACDL_TAG_NAMING SKIPPED record.
* scripts/run_ci.sh: lint stage compiles the new adapter modules.

202 tests pass; CI pipeline OK (lint + test + check-only).

Deviations:
- Wiz adapt() had an AttributeError on bare-list top-level input
  (data.get() on a list); fixed to dispatch on isinstance(data, list)
  before calling .get(). No spec change — bare-list handling is implied
  by the original docstring's "data if isinstance(data, list)" branch.
- Kyverno _to_pcr({}) defaults result to "skipped" (entry.get("result",
  "skip") -> "skip"), not "error"; test expectation corrected. Added an
  explicit unknown-result-string test to cover the "error" fallback.

---ci---
project: acdl
phase: 23
milestone: v1.7
status: execute
---/ci---
This commit is contained in:
Jon Chery
2026-07-22 20:00:46 +00:00
parent dca35c78ec
commit 1fd37a2843
22 changed files with 935 additions and 31 deletions
+68
View File
@@ -0,0 +1,68 @@
# Kyverno Adapter
The Kyverno adapter translates Kyverno `PolicyReport` results to the
normalized ACDL
[`PolicyCheckResult`](../../schemas/policy_check_result.schema.json) schema
(engine: `"kyverno"`), mirroring the Checkov/Wiz adapter pattern.
## What Kyverno is
[Kyverno](https://kyverno.io/) is a Kubernetes-native policy engine. It
runs as an admission controller inside a cluster, validates / mutates /
generates K8s resources against declarative `ClusterPolicy` rules, and
publishes results to `PolicyReport` resources.
## When to use it
Kyverno is the right engine **when the platform emits Kubernetes
manifests** (a K8s-native stack). The ACDL platform today emits Terraform
only (D-053), so this adapter is **ready but inactive**: it ships now so
the schema path, severity/result mapping and sample policies are in place
ahead of the GitOps reconciler that will emit K8s manifests (roadmap).
## How the adapter translates PolicyReport results
`kyverno_adapter.py <policyreport.json> <contract-id>` reads a JSON file
containing a Kyverno `PolicyReport` (or just its `.results[]` array) and
emits a list of `PolicyCheckResult` dicts:
| Kyverno PolicyReport result field | PolicyCheckResult field |
|-----------------------------------|-------------------------|
| `policy` | `ruleId` (default `KYVERNO_UNKNOWN`) |
| `severity` | `severity` (lower-cased, mapped) |
| `result` | `result` (`pass`/`fail`/`error` as-is, `warn`/`skip``skipped`) |
| `message` | `message` |
| `resource` | `resourceRef` + `evidence.resource` |
| `namespace`, `kind`, `name` | `evidence.*` |
The adapter is read-only against a local JSON fixture; the GitOps
reconciler is responsible for fetching the live `PolicyReport` and writing
the file. When there are zero results, the adapter returns an empty list
(unlike Wiz it does not synthesize a SKIPPED record — Kyverno not running
is a deployment state, not a configuration gap).
## Roadmap dependency
This adapter activates when the GitOps reconciler (roadmap) emits K8s
manifests. Until then it is documentation-only; the pipeline does not
invoke it. The `engine: "kyverno"` enum value is present in
`schemas/policy_check_result.schema.json` so future records validate.
## Sample policies
The `policies/` directory holds three valid Kyverno `ClusterPolicy`
manifests (documentation-only today — the platform does not run them):
- `disallow-privileged-containers.yaml` — fail pods with
`securityContext.privileged: true`.
- `require-resource-labels.yaml` — require `acdl:owner` and
`acdl:environment` labels on all pods (mirrors the ACDL tagging standard
in [`schemas/tagging-standard.json`](../../schemas/tagging-standard.json)).
- `require-image-digests.yaml` — require container images to reference a
digest (`image@sha256:...`), not a mutable tag.
## Schema path
The output records validate against
[`schemas/policy_check_result.schema.json`](../../schemas/policy_check_result.schema.json)
(`engine: "kyverno"` was already in the enum and is retained in Phase 23).
View File
+81
View File
@@ -0,0 +1,81 @@
"""Kyverno adapter — translate Kyverno PolicyReport results to ACDL PolicyCheckResult records.
Kyverno is a Kubernetes-native policy engine. It evaluates K8s manifests
and produces PolicyReport resources. This adapter translates those results
to the normalized PolicyCheckResult schema (engine: "kyverno").
D-053: the platform emits Terraform, not K8s manifests. This adapter is
ready but inactive for Terraform-only stacks. It activates when the GitOps
reconciler (roadmap) emits K8s manifests. Sample policies are included as
documentation at adapters/kyverno/policies/.
CLI: kyverno_adapter.py <policyreport.json> <contract-id>
"""
import datetime
import json
import sys
SEVERITY_MAP = {
"critical": "critical",
"high": "high",
"medium": "medium",
"low": "low",
"info": "info",
}
RESULT_MAP = {
"pass": "pass",
"fail": "fail",
"warn": "skipped",
"error": "error",
"skip": "skipped",
}
def _iso8601_now():
return datetime.datetime.now(datetime.timezone.utc).strftime("%Y-%m-%dT%H:%M:%SZ")
def _to_pcr(entry, contract_id):
severity_raw = entry.get("severity", "info")
severity = SEVERITY_MAP.get(str(severity_raw).lower(), "info")
result_raw = entry.get("result", "skip")
result = RESULT_MAP.get(str(result_raw).lower(), "error")
return {
"contractId": contract_id,
"evaluatedAt": _iso8601_now(),
"engine": "kyverno",
"ruleId": entry.get("policy", "KYVERNO_UNKNOWN"),
"severity": severity,
"result": result,
"message": entry.get("message", ""),
"evidence": {
"resource": entry.get("resource", ""),
"namespace": entry.get("namespace", ""),
"kind": entry.get("kind", ""),
"name": entry.get("name", ""),
},
"resourceRef": entry.get("resource", ""),
}
def adapt(policyreport_json_path, contract_id):
with open(policyreport_json_path, "r", encoding="utf-8") as fh:
data = json.load(fh)
out = []
# Kyverno PolicyReport has a .results[] array
results = data.get("results", [])
if not isinstance(results, list):
results = []
for entry in results:
out.append(_to_pcr(entry, contract_id))
return out
if __name__ == "__main__":
if len(sys.argv) != 3:
print("usage: kyverno_adapter.py <policyreport.json> <contract-id>", file=sys.stderr)
sys.exit(2)
print(json.dumps(adapt(sys.argv[1], sys.argv[2]), indent=2))
@@ -0,0 +1,27 @@
apiVersion: kyverno.io/v1
kind: ClusterPolicy
metadata:
name: disallow-privileged-containers
annotations:
policies.kyverno.io/title: Disallow Privileged Containers
policies.kyverno.io/category: Security
policies.kyverno.io/severity: high
policies.kyverno.io/subject: Pod
spec:
validationFailureAction: audit
background: true
rules:
- name: require-non-privileged
match:
any:
- resources:
kinds:
- Pod
validate:
message: "Privileged containers are not allowed. Set securityContext.privileged to false."
pattern:
spec:
containers:
- name: "*"
securityContext:
privileged: "false"
@@ -0,0 +1,26 @@
apiVersion: kyverno.io/v1
kind: ClusterPolicy
metadata:
name: require-image-digests
annotations:
policies.kyverno.io/title: Require Image Digests
policies.kyverno.io/category: Supply Chain
policies.kyverno.io/severity: high
policies.kyverno.io/subject: Pod
spec:
validationFailureAction: audit
background: true
rules:
- name: require-digest-reference
match:
any:
- resources:
kinds:
- Pod
validate:
message: "Container images must reference a digest (e.g. image@sha256:...), not a mutable tag."
pattern:
spec:
containers:
- name: "*"
image: "*@sha256:*"
@@ -0,0 +1,37 @@
apiVersion: kyverno.io/v1
kind: ClusterPolicy
metadata:
name: require-resource-labels
annotations:
policies.kyverno.io/title: Require ACDL Resource Labels
policies.kyverno.io/category: Governance
policies.kyverno.io/severity: medium
policies.kyverno.io/subject: Pod
spec:
validationFailureAction: audit
background: true
rules:
- name: require-acdl-owner-label
match:
any:
- resources:
kinds:
- Pod
validate:
message: "Pods must carry the acdl:owner label (ACDL tagging standard)."
pattern:
metadata:
labels:
acdl:owner: "?*"
- name: require-acdl-environment-label
match:
any:
- resources:
kinds:
- Pod
validate:
message: "Pods must carry the acdl:environment label (ACDL tagging standard)."
pattern:
metadata:
labels:
acdl:environment: "?*"
+8 -17
View File
@@ -6,8 +6,10 @@ schemas/policy_check_result.schema.json. Run Checkov with --soft-fail so
Checkov never exits non-zero; the confidence signal decides the gate, not
Checkov's exit code.
Spike scope (D-043): tag/naming is a single SKIPPED record. A custom
Checkov YAML rule for tag presence lands in v1.2.
The ACDL tagging standard (D-054, D-043 closure) is enforced by a custom
Checkov rule at adapters/terraform/policy/custom_rules/acdl_tagging.py,
loaded via --external-checks-dir. The adapter therefore maps
ACDL_TAG_NAMING as a real rule (no synthetic SKIPPED record is emitted).
"""
import datetime
@@ -27,6 +29,10 @@ RULE_MAP = {
"CKV_AWS_40": ("iam-wildcard", "medium"),
"CKV_AWS_7": ("kms-key-reference", "medium"),
"CKV_AWS_33": ("kms-key-reference", "medium"),
# D-054 / D-043 closure: ACDL_TAG_NAMING is now a real custom Checkov
# rule (adapters/terraform/policy/custom_rules/acdl_tagging.py), loaded
# via --external-checks-dir. No synthetic SKIPPED record is emitted.
"ACDL_TAG_NAMING": ("tagging-standard", "medium"),
}
_RESULT_MAP = {"PASSED": "pass", "FAILED": "fail", "SKIPPED": "skipped"}
@@ -60,20 +66,6 @@ def _to_pcr(checkov_record, contract_id, result_str):
}
def _emit_tag_naming_skipped(contract_id):
return {
"contractId": contract_id,
"evaluatedAt": _iso8601_now(),
"engine": "checkov",
"ruleId": "ACDL_TAG_NAMING",
"severity": "info",
"result": "skipped",
"message": "tag/naming check deferred to v1.2 (D-043)",
"evidence": {},
"resourceRef": "",
}
def adapt(checkov_json_path, contract_id):
with open(checkov_json_path, "r", encoding="utf-8") as fh:
data = json.load(fh)
@@ -88,7 +80,6 @@ def adapt(checkov_json_path, contract_id):
out.append(_to_pcr(rec, contract_id, "FAILED"))
for rec in results.get("skipped_checks", []):
out.append(_to_pcr(rec, contract_id, "SKIPPED"))
out.append(_emit_tag_naming_skipped(contract_id))
return out
@@ -0,0 +1,34 @@
# ACDL Custom Checkov Rules
This directory holds ACDL-authored Checkov custom rules, written in the
[Checkov Python custom-rule framework](https://www.checkov.io/4.Contributing/Custom%20Policies.html).
## Files
- `acdl_tagging.py``ACDL_TAG_NAMING` (D-054): ensures every taggable AWS
resource carries the four required ACDL tags
(`acdl:owner`, `acdl:contract`, `acdl:environment`, `acdl:cost-center`).
This rule replaces the synthetic SKIPPED `ACDL_TAG_NAMING` record that the
Checkov adapter previously emitted (D-043 closure). The canonical tag set
is declared in [`schemas/tagging-standard.json`](../../../schemas/tagging-standard.json).
## How Checkov loads them
Checkov custom rules are discovered via the `--external-checks-dir` flag.
`scripts/run_platform.sh` invokes Checkov with:
```
checkov -f terraform/spike/main.tf --framework terraform -o json --soft-fail \
--external-checks-dir adapters/terraform/policy/custom_rules/
```
Checkov imports each `*.py` file in the directory and instantiates the
module-level `check` object (see the `check = AcdlTaggingStandard()` line at
the bottom of `acdl_tagging.py`).
## Severity / result mapping
The Checkov adapter (`adapters/terraform/policy/checkov_adapter.py`)
maps `ACDL_TAG_NAMING` to `(tagging-standard, medium)` in `RULE_MAP`. The
custom rule therefore produces real `PASS`/`FAIL` PolicyCheckResult records,
feeding the confidence signal instead of the old SKIPPED placeholder.
@@ -0,0 +1,55 @@
"""ACDL tagging standard custom Checkov rule (D-054).
Checks that all taggable AWS resources have the required ACDL tags:
acdl:owner, acdl:contract, acdl:environment, acdl:cost-center
Fails (severity medium) when any required tag is missing.
Closes the D-043 deferral (the SKIPPED ACDL_TAG_NAMING placeholder
becomes a real check).
"""
from __future__ import annotations
from checkov.terraform.checks.resource.base_resource_check import BaseResourceCheck
from checkov.common.models.enums import CheckResult, CheckCategories
from checkov.common.models.consts import graph_resource_name_utils
REQUIRED_TAGS = ("acdl:owner", "acdl:contract", "acdl:environment", "acdl:cost-center")
# Resources that support tags (exclude resources that have no tags attribute)
NON_TAGGABLE_TYPES = (
"aws_cloudfront_origin_access_control",
"aws_lambda_function_url",
"aws_route_table_association",
"aws_internet_gateway",
)
class AcdlTaggingStandard(BaseResourceCheck):
def __init__(self):
name = "Ensure all taggable AWS resources have required ACDL tags"
check_id = "ACDL_TAG_NAMING"
supported_resources = ["*"] # all resources
categories = [CheckCategories.GENERAL_SECURITY]
super().__init__(name=name, check_id=check_id, categories=categories, supported_resources=supported_resources)
def scan_resource_conf(self, conf, entity_type):
# Skip non-taggable resources
if entity_type in NON_TAGGABLE_TYPES:
return CheckResult.PASSED
# Check for a tags block
tags = conf.get("tags")
if not tags:
return CheckResult.FAILED
tag_keys = set()
if isinstance(tags, list) and tags:
tag_block = tags[0]
if isinstance(tag_block, dict):
tag_keys = set(tag_block.keys())
elif isinstance(tags, dict):
tag_keys = set(tags.keys())
missing = [t for t in REQUIRED_TAGS if t not in tag_keys]
if missing:
return CheckResult.FAILED
return CheckResult.PASSED
check = AcdlTaggingStandard()
+55
View File
@@ -0,0 +1,55 @@
# Wiz Adapter
The Wiz adapter translates Wiz API issue records to the normalized ACDL
[`PolicyCheckResult`](../../schemas/policy_check_result.schema.json) schema
(engine: `"wiz"`), mirroring the Checkov adapter pattern.
## What Wiz is
[Wiz](https://www.wiz.io/) is a cloud security SaaS platform that
continuously scans CSPM / CWPP / KSPM findings across AWS, Azure, GCP and
Kubernetes. It exposes a GraphQL/REST API for fetching issue records.
## Adapter behaviour
`wiz_adapter.py <wiz_issues.json> <contract-id>` reads a JSON file of Wiz
issue records (the shape returned by the Wiz `issues` GraphQL query /
list endpoint) and emits a list of `PolicyCheckResult` dicts:
| Wiz field | PolicyCheckResult field |
|------------------|------------------------------------------------------------|
| `id` / `control.id` | `ruleId` |
| `severity` | `severity` (mapped `CRITICAL/HIGH/MEDIUM/LOW/INFO`) |
| `status` | `result` (`OPEN→fail`, `RESOLVED→pass`, `IN_PROGRESS/DISMISSED→skipped`) |
| `title` / `control.name` | `message` |
| `entity.id` | `resourceRef` + `evidence.resource` |
| `entity.{name,cloudPlatform,subscriptionId}` | `evidence.*` |
The adapter is read-only against a local JSON fixture; the pipeline is
responsible for fetching from Wiz (when configured) and writing the file.
## Offline / degraded behaviour (D-052)
When Wiz is not configured the pipeline passes an empty issues payload (or
simply does not invoke the adapter). The adapter degrades gracefully:
- an empty `issues` list → the adapter emits a single `WIZ_NOT_CONFIGURED`
`PolicyCheckResult` with `result: "skipped"` so the confidence policy
input stays non-empty (and does not falsely inflate the score).
`is_configured()` returns `True` only when the `WIZ_API_TOKEN`
environment variable is set; the pipeline uses it to decide whether to
fetch and invoke the adapter at all.
## Configuration
| Env var | Required | Purpose |
|-----------------|----------|--------------------------------------------------|
| `WIZ_API_TOKEN` | yes | Bearer token for the Wiz REST API. When unset, `is_configured()` returns `False`. |
| `WIZ_ENDPOINT` | no | Wiz API endpoint (defaults to `https://api.wiz.io` when implemented). |
## Schema path
The output records validate against
[`schemas/policy_check_result.schema.json`](../../schemas/policy_check_result.schema.json)
(`engine: "wiz"` was added to the enum in Phase 23).
View File
+106
View File
@@ -0,0 +1,106 @@
"""Wiz adapter — translate Wiz API results to ACDL PolicyCheckResult records.
Wiz is a SaaS security platform with a REST API (issues, security graph
queries). This adapter translates Wiz issue records to the normalized
PolicyCheckResult schema (engine: "wiz"), matching the Checkov adapter
pattern.
D-052: stub + schema path. The adapter degrades gracefully when Wiz is
not configured — it emits a single SKIPPED record (WIZ_NOT_CONFIGURED)
so the confidence policy input stays non-empty. The pipeline invokes it
optionally when WIZ_API_TOKEN is set.
CLI: wiz_adapter.py <wiz_issues.json> <contract-id>
"""
import datetime
import json
import os
import sys
SEVERITY_MAP = {
"CRITICAL": "critical",
"HIGH": "high",
"MEDIUM": "medium",
"LOW": "low",
"INFO": "info",
}
RESULT_MAP = {
"OPEN": "fail",
"RESOLVED": "pass",
"IN_PROGRESS": "skipped",
"DISMISSED": "skipped",
}
def _iso8601_now():
return datetime.datetime.now(datetime.timezone.utc).strftime("%Y-%m-%dT%H:%M:%SZ")
def _to_pcr(wiz_issue, contract_id):
severity_raw = wiz_issue.get("severity", "INFO")
severity = SEVERITY_MAP.get(str(severity_raw).upper(), "info")
status = wiz_issue.get("status", "OPEN")
result = RESULT_MAP.get(str(status).upper(), "error")
control = wiz_issue.get("control", {})
return {
"contractId": contract_id,
"evaluatedAt": _iso8601_now(),
"engine": "wiz",
"ruleId": wiz_issue.get("id", control.get("id", "WIZ_UNKNOWN")),
"severity": severity,
"result": result,
"message": wiz_issue.get("title", control.get("name", "")),
"evidence": {
"resource": wiz_issue.get("entity", {}).get("id"),
"resource_name": wiz_issue.get("entity", {}).get("name"),
"cloud_platform": wiz_issue.get("entity", {}).get("cloudPlatform"),
"subscription_id": wiz_issue.get("entity", {}).get("subscriptionId"),
},
"resourceRef": wiz_issue.get("entity", {}).get("id", ""),
}
def _emit_not_configured(contract_id):
return {
"contractId": contract_id,
"evaluatedAt": _iso8601_now(),
"engine": "wiz",
"ruleId": "WIZ_NOT_CONFIGURED",
"severity": "info",
"result": "skipped",
"message": "Wiz adapter not configured (WIZ_API_TOKEN not set); degraded gracefully (D-052).",
"evidence": {},
"resourceRef": "",
}
def adapt(wiz_json_path, contract_id):
with open(wiz_json_path, "r", encoding="utf-8") as fh:
data = json.load(fh)
out = []
# Accept either a bare list of issues or an object with an "issues" key.
if isinstance(data, list):
issues = data
else:
issues = data.get("issues", [])
if not isinstance(issues, list):
issues = []
for issue in issues:
out.append(_to_pcr(issue, contract_id))
if not out:
out.append(_emit_not_configured(contract_id))
return out
def is_configured():
return bool(os.environ.get("WIZ_API_TOKEN"))
if __name__ == "__main__":
if len(sys.argv) != 3:
print("usage: wiz_adapter.py <wiz_issues.json> <contract-id>", file=sys.stderr)
sys.exit(2)
print(json.dumps(adapt(sys.argv[1], sys.argv[2]), indent=2))