aero-agent-roles 1.0.0

package/roles/as9100-quality-auditor/templates/findings-report-template.md

@@ -0,0 +1,119 @@
+# AS9100 Internal Audit Findings Report
+
+**Supplier:** AeroForge Precision Machining, LLC
+**Audit date:** 2026-09-05
+**Auditor:** M. Reyes (independent of area owner J. Torres: yes)
+**Status:** draft-for-review
+
+## Audit plan
+
+- Audit scope: CNC machining and production control (8.5.1), inspection and test (8.6), calibration control (7.1.5), nonconformance disposition (8.7), corrective action (10.2).
+- Criteria: AS9100D clauses (6.1, 7.1.5, 8.1.2, 8.6, 8.7, 10.2), customer requirements flow-down, PO-7712.
+- Audit team: M. Reyes (lead auditor).
+- Process risk category: high; next audit due 2027-03-05 (risk-scaled 12-month base interval).
+- Records sampled: 20 of 400 records sampled at 0.95 confidence (square-root rule).
+
+## Findings summary
+
+- Major nonconformities: 2
+- Minor nonconformities: 2
+- Observations (opportunities for improvement): 1
+- Total findings: 5
+- AS9100D clauses cited: 10.2, 6.1, 7.1.5, 8.1.2, 8.7.
+
+## Findings report
+
+### Nonconformities
+
+| NC ID | Clause | Classification | Objective evidence | Requirement |
+|---|---|---|---|---|
+| NC-01 | 8.7 | major | NCR-2214: bracket P/N BN-4470-2 (safety-critical, flight control linkage) dispositioned use-as-is by the supplier MRB; no customer waiver on file and the disposition record has no customer-approval entry. Twelve brackets of the same lot shipped under delivery DLV-3390. | Nonconforming output must be identified, segregated, and dispositioned by an authorized authority; use-as-is outside the original specification requires customer approval. |
+| NC-02 | 10.2 | major | NCR-2210/2211/2214 record the same oversize-bore characteristic within six months. CAPA-118 was closed with effectiveness evidence restating the root cause rather than an observed result. | On nonconformity, the organization must react, evaluate the need for action to eliminate the cause so it does not recur, and verify the effectiveness of the action taken. |
+| NC-03 | 7.1.5 | minor | Torque wrench TQ-009 in the assembly tool crib carried a calibration sticker due 2026-08-30; the crib sign-out log shows no issue after the sticker expiry date. | Monitoring and measuring resources must be calibrated or verified at defined intervals; suspect measurements must be assessed for validity. |
+| NC-04 | 8.1.2 | minor | ECO-4821 (drawing BN-4470 rev A to rev B) approved 2026-08-10; the configuration baseline register entry for BN-4470 still shows rev A as of the audit date. No nonconforming product resulted. | Configuration management must establish and maintain a configuration baseline and control changes to it. |
+| NC-05 | 6.1 | observation | The operational risk register is complete and current; mitigation actions for the coating process change are recorded but no residual RPN re-score follows the recorded reductions. | No violation found; the register meets clause 6.1. Re-scoring residual risk closes the mitigation loop demonstrated elsewhere in the QMS. |
+
+### Finding details
+
+#### NC-01 - MAJOR (clause 8.7)
+
+**Title:** Use-as-is disposition on a safety-critical part without customer approval
+**Classification rationale:** severity impact 5 is 4-5, which is a major nonconformity.
+**Objective evidence:** NCR-2214: bracket P/N BN-4470-2 (safety-critical, flight control linkage) dispositioned use-as-is by the supplier MRB; no customer waiver on file and the disposition record has no customer-approval entry. Twelve brackets of the same lot shipped under delivery DLV-3390.
+**Requirement violated:** Nonconforming output must be identified, segregated, and dispositioned by an authorized authority; use-as-is outside the original specification requires customer approval.
+
+Corrective-action record (closure chain status: effectiveness-pending):
+- Containment: Quarantine the twelve shipped brackets of lot 2214B at the customer site; suspend MRB use-as-is disposition until the procedure is corrected.
+- Root cause: The MRB disposition procedure has no customer-approval step for use-as-is > The procedure was written when the MRB handled non-aerospace parts only > Customer approval flow was never added when the aerospace order book opened > No checklist gate sits between the disposition decision and product release
+- Corrective action: Revise the disposition procedure to require customer approval before any use-as-is release and add a release checklist gate; retrain the MRB board.
+- Effectiveness evidence: pending - not yet recorded
+
+#### NC-02 - MAJOR (clause 10.2)
+
+**Title:** Recurring nonconformity with no root cause or effectiveness verification
+**Classification rationale:** severity impact 3 with systemic spread, which escalates a minor to a major nonconformity.
+**Objective evidence:** NCR-2210/2211/2214 record the same oversize-bore characteristic within six months. CAPA-118 was closed with effectiveness evidence restating the root cause rather than an observed result.
+**Requirement violated:** On nonconformity, the organization must react, evaluate the need for action to eliminate the cause so it does not recur, and verify the effectiveness of the action taken.
+
+Corrective-action record (closure chain status: effectiveness-pending):
+- Containment: 100% bore inspection of the affected part numbers in stock and in work.
+- Root cause: Tool wear compensation is not verified between operator shifts > The compensation check is a manual step with no scheduled trigger > The CNC program pauses at shift change but no measurement is required before restart
+- Corrective action: Add an in-process bore measurement gate at every shift change, tied to the SPC record for the characteristic.
+- Effectiveness evidence: pending - not yet recorded
+
+#### NC-03 - MINOR (clause 7.1.5)
+
+**Title:** Calibration sticker expired on a torque wrench in the tool crib
+**Classification rationale:** severity impact 2 is 2-3 with no systemic spread, which is a minor nonconformity.
+**Objective evidence:** Torque wrench TQ-009 in the assembly tool crib carried a calibration sticker due 2026-08-30; the crib sign-out log shows no issue after the sticker expiry date.
+**Requirement violated:** Monitoring and measuring resources must be calibrated or verified at defined intervals; suspect measurements must be assessed for validity.
+
+Corrective-action record (closure chain status: effectiveness-pending):
+- Containment: Remove TQ-009 to calibration; verify the sign-out log to bound the period of possible use.
+- Root cause: No recall alert fired when the interval expired > Calibration due dates are tracked on a paper log reviewed monthly > The monthly review ran two weeks late
+- Corrective action: Move calibration due-date tracking to the shop system with an automatic recall alert at interval expiry.
+- Effectiveness evidence: pending - not yet recorded
+
+#### NC-04 - MINOR (clause 8.1.2)
+
+**Title:** Configuration baseline register not updated for approved engineering change ECO-4821
+**Classification rationale:** severity impact 3 is 2-3 with no systemic spread, which is a minor nonconformity.
+**Objective evidence:** ECO-4821 (drawing BN-4470 rev A to rev B) approved 2026-08-10; the configuration baseline register entry for BN-4470 still shows rev A as of the audit date. No nonconforming product resulted.
+**Requirement violated:** Configuration management must establish and maintain a configuration baseline and control changes to it.
+
+Corrective-action record (closure chain status: effectiveness-pending):
+- Containment: Update the baseline register for BN-4470 rev B; audit open ECOs against the register.
+- Root cause: Baseline register updates are batched monthly > No trigger links ECO approval to the register update > The register owner has no change-approval notification
+- Corrective action: Link ECO approval records to the baseline register so approved changes update the register at approval time.
+- Effectiveness evidence: pending - not yet recorded
+
+#### NC-05 - OBSERVATION (clause 6.1)
+
+**Title:** Opportunity: risk register entries lack post-mitigation residual scoring
+**Classification rationale:** severity impact 1 is an opportunity for improvement, reported as an observation.
+**Objective evidence:** The operational risk register is complete and current; mitigation actions for the coating process change are recorded but no residual RPN re-score follows the recorded reductions.
+**Requirement violated:** No violation found; the register meets clause 6.1. Re-scoring residual risk closes the mitigation loop demonstrated elsewhere in the QMS.
+
+No corrective action required (observation, clause 6.1); improvement tracked separately.
+
+### Observations (no violation found, improvement note)
+
+- NC-05 (clause 6.1): Opportunity: risk register entries lack post-mitigation residual scoring. The operational risk register is complete and current; mitigation actions for the coating process change are recorded but no residual RPN re-score follows the recorded reductions.
+
+### Corrective-action follow-up
+
+| NC ID | Root cause | Containment | Corrective action | Verification | Status |
+|---|---|---|---|---|---|
+| NC-01 | use-as-is release without customer approval on safety-critical parts | Quarantine the twelve shipped brackets of lot 2214B at the customer site; suspend MRB use-as-is disposition until the procedure is corrected. | Revise the disposition procedure to require customer approval before any use-as-is release and add a release checklist gate; retrain the MRB board. | pending - effectiveness not yet verified | effectiveness-pending |
+| NC-02 | oversize-bore recurrence from unverified tool-wear compensation at shift change | 100% bore inspection of the affected part numbers in stock and in work. | Add an in-process bore measurement gate at every shift change, tied to the SPC record for the characteristic. | pending - effectiveness not yet verified | effectiveness-pending |
+| NC-03 | expired calibration sticker from a late monthly due-date review | Remove TQ-009 to calibration; verify the sign-out log to bound the period of possible use. | Move calibration due-date tracking to the shop system with an automatic recall alert at interval expiry. | pending - effectiveness not yet verified | effectiveness-pending |
+| NC-04 | baseline register updated monthly instead of at ECO approval | Update the baseline register for BN-4470 rev B; audit open ECOs against the register. | Link ECO approval records to the baseline register so approved changes update the register at approval time. | pending - effectiveness not yet verified | effectiveness-pending |
+| NC-05 | n/a - observation | n/a - observation | n/a - observation | n/a - observation | observation |
+
+## Closure
+
+- Closure recommendation: PROPOSED - the QA manager signs audit closure; NC-01/02 effectiveness verification at the follow-up audit is required before any closure is recorded.
+- No finding is recorded closed: closure requires the corrective action taken, the root cause statement and the effectiveness check all on file (verification gate).
+
+---
+*DRAFT - audit evidence report generated by Aero Agent Roles as9100-quality-auditor core (2026-09-05) for human QA review. Not a certification, supplier-approval, or closure decision; the QA manager signs closure.*

package/roles/as9100-quality-auditor/tests/test_as9100_cli_profile.py

@@ -0,0 +1,105 @@
+#!/usr/bin/env python3
+"""CLI test: as9100-quality-auditor build --profile (PROFILE-SCHEMA.md v1).
+
+Verifies the --profile path end-to-end through cli.py (offline, no
+network): a program profile (audited supplier context) prepends the
+'## Program context' header to the findings report and records
+profile.customer in the evidence model.json, exactly like the
+do178c-cert-engineer reference. Also checks the no-profile path is
+unchanged and a malformed profile fails loudly with exit 1.
+"""
+import json
+import os
+import shutil
+import subprocess
+import sys
+import tempfile
+import unittest
+
+ROLE_DIR = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
+CLI = os.path.join(ROLE_DIR, "cli.py")
+# repo root is 4 up from tests/ (tests -> role -> roles -> repo)
+REPO_ROOT = os.path.dirname(os.path.dirname(os.path.dirname(
+    os.path.dirname(os.path.abspath(__file__)))))
+PROFILE = os.path.join(REPO_ROOT, "profiles", "example-airframer.json")
+EXPECTED_CUSTOMER = "Example Airframers Ltd"
+
+
+def run_build(outdir, extra_args=None):
+    args = [sys.executable, CLI, "build", "--out",
+            os.path.join(outdir, "report.md"), "--bundle"]
+    if extra_args:
+        args.extend(extra_args)
+    return subprocess.run(args, capture_output=True, text=True)
+
+
+class TestAs9100CliProfile(unittest.TestCase):
+
+    def setUp(self):
+        self.tmp = tempfile.mkdtemp()
+        if not os.path.exists(PROFILE):
+            self.skipTest("profiles/example-airframer.json not present")
+
+    def tearDown(self):
+        shutil.rmtree(self.tmp, ignore_errors=True)
+
+    def _model(self):
+        return json.load(open(os.path.join(self.tmp, "evidence",
+                                           "model.json")))
+
+    def test_build_with_profile_exits_zero_header_first(self):
+        r = run_build(self.tmp, ["--profile", PROFILE])
+        self.assertEqual(r.returncode, 0, r.stderr[-500:])
+        md = open(os.path.join(self.tmp, "report.md")).read()
+        self.assertTrue(md.startswith("## Program context"),
+                        "profile context header must lead the report")
+        self.assertIn("Customer: " + EXPECTED_CUSTOMER, md)
+        self.assertIn("profile: " + EXPECTED_CUSTOMER, r.stdout)
+
+    def test_model_json_records_profile_customer(self):
+        r = run_build(self.tmp, ["--profile", PROFILE])
+        self.assertEqual(r.returncode, 0)
+        m = self._model()
+        self.assertEqual(m["role"], "as9100-quality-auditor")
+        self.assertIn("profile", m)
+        self.assertEqual(m["profile"]["customer"], EXPECTED_CUSTOMER)
+        self.assertEqual(m["profile"]["basis"], "FAR-25")
+        self.assertEqual(m["profile"]["authority"], "FAA")
+        # profile tailors context; the report body is still the audit
+        self.assertEqual(m["status"], "draft-for-review")
+        self.assertTrue(m["supplier"])
+
+    def test_profiled_report_still_passes_gates(self):
+        r = run_build(self.tmp, ["--profile", PROFILE])
+        self.assertEqual(r.returncode, 0)
+        g = json.load(open(os.path.join(self.tmp, "evidence",
+                                        "gates.json")))
+        self.assertTrue(g["all_pass"])
+        self.assertEqual(g["exit_code"], 0)
+        # gate-check command also passes on the profiled markdown
+        c = subprocess.run(
+            [sys.executable, CLI, "check", "--file",
+             os.path.join(self.tmp, "report.md")],
+            capture_output=True, text=True)
+        self.assertEqual(c.returncode, 0, c.stdout)
+        self.assertIn("RESULT: PASS", c.stdout)
+
+    def test_no_profile_unchanged(self):
+        r = run_build(self.tmp)
+        self.assertEqual(r.returncode, 0, r.stderr[-500:])
+        md = open(os.path.join(self.tmp, "report.md")).read()
+        self.assertTrue(md.startswith("# AS9100 Internal Audit "
+                                      "Findings Report"))
+        self.assertNotIn("profile", self._model())
+
+    def test_malformed_profile_fails_loudly(self):
+        bad = os.path.join(self.tmp, "bad-profile.json")
+        with open(bad, "w") as f:
+            json.dump({"program": "no customer here"}, f)
+        r = run_build(self.tmp, ["--profile", bad])
+        self.assertEqual(r.returncode, 1)
+        self.assertIn("error: bad profile", r.stdout)
+
+
+if __name__ == "__main__":
+    unittest.main()

package/roles/as9100-quality-auditor/tests/test_as9100_core.py

@@ -0,0 +1,204 @@
+#!/usr/bin/env python3
+"""Test as9100_core: the executable engine of the AS9100 auditor role.
+
+Proves the role can DO its job standalone (no AeroSkills needed):
+NC classification on the finding ladder, corrective action closure
+chain scoring, audit program numbers (sample size, due date), findings
+report generation, and evidence-gate checks.
+"""
+import os
+import sys
+import unittest
+
+sys.path.insert(0, os.path.join(os.path.dirname(os.path.abspath(__file__)), "..", "core"))
+from as9100_core import (
+    AS9100D_SECTIONS, BASE_INTERVAL_MONTHS, CLAUSE_BY_ID,
+    NC_CLASSIFICATIONS, audit_due_date, audit_findings, audit_sample_size,
+    build_findings, check_findings, check_findings_markdown,
+    classify_nc, containment_ok, corrective_action_status,
+    effectiveness_evidence_ok, example_findings_markdown, example_item,
+    render_findings_markdown, root_cause_chain_ok,
+)
+
+
+class TestClassificationLadder(unittest.TestCase):
+
+    def test_high_severity_is_major(self):
+        self.assertEqual(classify_nc(4, False, False), "major")
+        self.assertEqual(classify_nc(5, False, False), "major")
+
+    def test_mid_severity_is_minor(self):
+        self.assertEqual(classify_nc(2, False, False), "minor")
+        self.assertEqual(classify_nc(3, False, False), "minor")
+
+    def test_systemic_escalates_minor_to_major(self):
+        self.assertEqual(classify_nc(2, True, False), "major")
+        self.assertEqual(classify_nc(3, True, False), "major")
+
+    def test_escaped_detection_escalates_to_major(self):
+        # Detection failure means containment of suspect output is
+        # required, which is a major regardless of severity.
+        self.assertEqual(classify_nc(1, False, True), "major")
+        self.assertEqual(classify_nc(2, False, True), "major")
+
+    def test_low_severity_is_observation(self):
+        self.assertEqual(classify_nc(1, False, False), "observation")
+
+    def test_severity_out_of_range_raises(self):
+        for bad in (0, 6, -1, 9):
+            with self.assertRaises(ValueError):
+                classify_nc(bad, False, False)
+
+
+class TestCorrectiveActionChain(unittest.TestCase):
+
+    def test_closure_chain_stages(self):
+        record = {
+            "problem": "recurring defect",
+            "containment": "quarantine suspect lot",
+            "whys": ["cause a", "cause b", "cause c"],
+            "corrective_action": "update process control",
+            "effectiveness_evidence": "",
+            "root_cause_statement": "cause c",
+        }
+        self.assertEqual(corrective_action_status(record),
+                         "effectiveness-pending")
+        record["effectiveness_evidence"] = "three clean lots observed"
+        self.assertEqual(corrective_action_status(record), "closed")
+
+    def test_circular_effectiveness_evidence_fails(self):
+        record = {
+            "problem": "recurring defect",
+            "containment": "quarantine",
+            "whys": ["a", "b", "c"],
+            "corrective_action": "update control",
+            "effectiveness_evidence": "cause c",
+            "root_cause_statement": "cause c",
+        }
+        self.assertFalse(effectiveness_evidence_ok(
+            record["effectiveness_evidence"],
+            record["root_cause_statement"]))
+        self.assertEqual(corrective_action_status(record),
+                         "effectiveness-pending")
+
+    def test_chain_helpers(self):
+        self.assertTrue(containment_ok("quarantine"))
+        self.assertFalse(containment_ok("none"))
+        self.assertTrue(root_cause_chain_ok(["a", "b", "c"]))
+        self.assertFalse(root_cause_chain_ok(["a", "b"]))
+        self.assertFalse(root_cause_chain_ok(["a", "a", "a"]))
+
+
+class TestAuditNumbers(unittest.TestCase):
+
+    def test_sample_size_at_confidence_anchors(self):
+        self.assertEqual(audit_sample_size(400, 0.95), 20)
+        self.assertEqual(audit_sample_size(400, 0.90), 16)
+        self.assertEqual(audit_sample_size(400, 0.99), 24)
+
+    def test_sample_size_validation(self):
+        with self.assertRaises(ValueError):
+            audit_sample_size(0)
+        with self.assertRaises(ValueError):
+            audit_sample_size(10, 0.4)
+
+    def test_audit_due_date_risk_scaling(self):
+        # 12-month base scaled by risk: high 0.5, low 1.5.
+        self.assertEqual(audit_due_date("2026-03-15", "high"), "2026-09-15")
+        self.assertEqual(audit_due_date("2026-03-15", "medium"), "2027-03-15")
+        self.assertEqual(audit_due_date("2026-03-15", "low"), "2027-09-15")
+
+    def test_audit_due_date_clamps_month_end(self):
+        self.assertEqual(audit_due_date("2026-01-31", "medium"), "2027-01-31")
+        self.assertEqual(audit_due_date("2024-01-31", "high"), "2024-07-31")
+
+
+class TestFindingsBuilder(unittest.TestCase):
+
+    def test_example_audit_covers_all_classifications(self):
+        item = example_item()
+        model = audit_findings(item)
+        cls = {f["classification"] for f in model["findings"]}
+        self.assertTrue({"major", "minor", "observation"} <= cls)
+        self.assertGreaterEqual(model["counts"]["major"], 1)
+        self.assertGreaterEqual(model["counts"]["minor"], 1)
+        self.assertGreaterEqual(model["counts"]["observation"], 1)
+        self.assertEqual(model["counts"]["total"], len(item.issues))
+
+    def test_findings_model_complete(self):
+        model = audit_findings(example_item())
+        self.assertEqual(model["document_type"],
+                         "AS9100 Internal Audit Findings Report")
+        self.assertIn("AS9100D", " ".join(model["audit"]["criteria"]))
+        for f in model["findings"]:
+            self.assertIn(f["clause"], CLAUSE_BY_ID)
+            self.assertIn(f["classification"], NC_CLASSIFICATIONS)
+            self.assertTrue(f["objective_evidence"])
+            self.assertTrue(f["requirement"])
+            if f["classification"] in ("major", "minor"):
+                ca = f["corrective_action"]
+                self.assertTrue(containment_ok(ca["containment"]))
+                self.assertTrue(root_cause_chain_ok(ca["whys"]))
+                self.assertTrue(ca["corrective_action"])
+                self.assertEqual(f["ca_status"], "effectiveness-pending")
+            else:
+                self.assertIsNone(f["corrective_action"])
+                self.assertEqual(f["ca_status"], "not-required")
+
+    def test_clause_ids_are_real(self):
+        # Example NCs cite only leaf-verified AS9100D clause ids.
+        model = audit_findings(example_item())
+        cited = {f["clause"] for f in model["findings"]}
+        self.assertTrue({"8.7", "10.2"} <= cited)
+        self.assertTrue(cited <= set(CLAUSE_BY_ID))
+        self.assertGreaterEqual(len(AS9100D_SECTIONS), 7)  # sections 4-10
+
+    def test_example_numbers_match_formulas(self):
+        item = example_item()
+        model = audit_findings(item)
+        self.assertEqual(model["sampling"]["sample_size"],
+                         audit_sample_size(item.record_population,
+                                           item.confidence_level))
+        self.assertEqual(model["sampling"]["sample_size"], 20)
+        self.assertEqual(model["audit"]["next_audit_due"],
+                         audit_due_date(item.audit_date, item.process_risk))
+
+    def test_build_findings_matches_audit_findings(self):
+        item = example_item()
+        self.assertEqual(build_findings(item), audit_findings(item))
+
+
+class TestGatesAndRender(unittest.TestCase):
+
+    def test_core_gates_all_pass(self):
+        model = audit_findings(example_item())
+        gates = check_findings(model)
+        self.assertTrue(gates["all_pass"], gates)
+
+    def test_markdown_gates_all_pass(self):
+        md = example_findings_markdown()
+        gates = check_findings_markdown(md)
+        self.assertTrue(gates["all_pass"], gates)
+
+    def test_render_has_no_blanks(self):
+        md = example_findings_markdown()
+        low = md.lower()
+        self.assertNotIn("___", md)
+        self.assertNotIn("| |", md)
+        self.assertNotIn("| - |", md)
+        self.assertIn("proposed", low)
+        self.assertIn("not a", low)
+
+    def test_standalone_no_skills_repo(self):
+        # Core must work with no AeroSkills present: set env to a bad path.
+        os.environ["AEROSKILLS_DEV"] = "/nonexistent"
+        item = example_item()
+        model = build_findings(item)
+        md = render_findings_markdown(model)
+        self.assertGreater(len(md), 1500)
+        self.assertTrue(check_findings(model)["all_pass"])
+        self.assertTrue(check_findings_markdown(md)["all_pass"])
+
+
+if __name__ == "__main__":
+    unittest.main()

package/roles/as9100-quality-auditor/tests/test_role_as9100.py

@@ -0,0 +1,90 @@
+#!/usr/bin/env python3
+"""Role test: AS9100 Quality / Internal Auditor.
+
+Offline verification: bound skills resolve, workflow order, findings
+template fields, boundaries, sources register.
+"""
+import os
+import unittest
+
+ROLES_REPO = os.path.dirname(os.path.dirname(os.path.dirname(os.path.dirname(
+    os.path.abspath(__file__)))))
+AEROSKILLS = os.environ.get("AEROSKILLS_DEV", os.path.expanduser("~/AeroSkills"))
+# Bound-skill resolution requires the Aero Agent Skills checkout (cross-repo
+# dev check); skip when absent so fresh public clones can still run make validate.
+HAS_SKILLS = os.path.isdir(os.path.join(AEROSKILLS, "skills"))
+ROLE_DIR = os.path.join(ROLES_REPO, "roles", "as9100-quality-auditor")
+FINDINGS = os.path.join(ROLE_DIR, "templates", "findings-report-template.md")
+
+EXPECTED_BOUND = [
+    "manufacturing-quality/as9100/internal-quality-audit",
+    "manufacturing-quality/as9100/document-control",
+    "manufacturing-quality/as9100/management-review",
+    "manufacturing-quality/as9100/nonconformance-control",
+    "manufacturing-quality/as9100/corrective-action",
+    "manufacturing-quality/as9100/risk-management",
+    "manufacturing-quality/as9100/supplier-control",
+    "manufacturing-quality/as9100/order-requirements-review",
+    "manufacturing-quality/as9100/calibration-control",
+    "manufacturing-quality/as9100/counterfeit-prevention",
+    "manufacturing-quality/as9100/fod-control",
+    "manufacturing-quality/as9100/quality",
+    "manufacturing-quality/as9102/first-article-inspection",
+    "manufacturing-quality/as9102/delta-fai",
+    "manufacturing-quality/as9102/fai-revalidation",
+]
+
+STAGES = ["Audit planning", "Document control check", "Management review check",
+          "Risk + ops review", "Nonconformance + corrective action",
+          "Supplier + purchasing", "Calibration + counterfeit + FOD",
+          "FAI (AS9102)", "Findings + report"]
+
+
+def role_text():
+    p = os.path.join(ROLE_DIR, "ROLE.md")
+    return open(p).read() if os.path.exists(p) else None
+
+
+class TestAs9100Role(unittest.TestCase):
+
+    def test_bound_skills_resolve(self):
+        if not HAS_SKILLS:
+            self.skipTest("Aero Agent Skills checkout not present (cross-repo dev check)")
+        for leaf in EXPECTED_BOUND:
+            self.assertTrue(
+                os.path.exists(os.path.join(AEROSKILLS, "skills", leaf, "SKILL.md")),
+                f"bound skill missing: {leaf}")
+
+    def test_stages_ordered(self):
+        role = role_text()
+        self.assertIsNotNone(role)
+        idx = [role.find(s) for s in STAGES]
+        self.assertTrue(all(i >= 0 for i in idx))
+        self.assertEqual(idx, sorted(idx), "stage order not deterministic")
+
+    def test_findings_fields_present(self):
+        if not os.path.exists(FINDINGS):
+            self.fail("findings-report-template.md missing")
+        text = open(FINDINGS).read()
+        for field in ["NC ID", "Clause", "Classification", "Objective evidence",
+                      "Root cause", "Corrective action", "Verification"]:
+            self.assertIn(field, text, f"field missing: {field}")
+        low = text.lower()
+        self.assertIn("proposed", low)
+        self.assertIn("not a", low)
+
+    def test_boundaries(self):
+        role = role_text()
+        self.assertIsNotNone(role)
+        for phrase in ["issue certification", "claim audit closure",
+                       "recommend supplier approval", "sign_off_required: true"]:
+            self.assertIn(phrase, role)
+
+    def test_sources_register(self):
+        self.assertTrue(
+            os.path.exists(os.path.join(ROLE_DIR, "SOURCES.md")),
+            "SOURCES.md missing")
+
+
+if __name__ == "__main__":
+    unittest.main()

package/roles/data-bus-avionics-engineer/ROLE.md

@@ -0,0 +1,146 @@
+---
+type: role
+name: data-bus-avionics-engineer
+title: "Data Bus / Avionics Network Engineer"
+status: draft
+domain: avionics
+deliverable_type: "avionics data bus loading + protocol assessment"
+standards_bound:
+  - id: arinc-429
+    tier: TIER-2
+    reference-only: true
+  - id: arinc-664
+    tier: TIER-2
+    reference-only: true
+  - id: mil-std-1553
+    tier: TIER-1
+    reference-only: false
+skills_bound:
+  - avionics/data-bus/arinc429-protocol
+  - avionics/data-bus/arinc429-bus-loading
+  - avionics/data-bus/arinc664-afdx
+  - avionics/data-bus/mil-std-1553
+  - avionics/data-bus/mil-std-1553-bus-loading
+tools_allowed: [stdlib, offline-file-processing]
+forbidden:
+  - "issue bus architecture sign-off"
+  - "claim network qualification approval"
+  - "assert a loading margin without the computed budget"
+  - "reproduce proprietary ARINC standard text"
+sign_off_required: true
+license: Apache-2.0
+compatibility: "agentskills.io ROLE.md; any SKILL.md host"
+metadata:
+  version: 0.1.0
+  author: ashfordeOU
+  skills_release: "v1.3.0+"
+---
+
+# Data Bus / Avionics Network Engineer
+
+## Role identity
+
+This role is the avionics data bus / network engineer for an aircraft
+program: it decodes ARINC 429 and MIL-STD-1553 words, budgets ARINC 429
+label-rate transmit schedules, computes MIL-STD-1553 bus-controller
+minor-frame loading, sizes ARINC 664 AFDX virtual links, and assembles
+the evidence that bus loading and protocol conformance gates are met.
+Use when a project must verify a data bus architecture against loading
+guidelines (ARINC 429 36 bit-times per word against the 80% design
+guideline; MIL-STD-1553 schedule time over the minor frame against the
+80% budget; AFDX VL bandwidth against the 100 Mbps link). Do NOT use for
+generic Ethernet networking (non-avionics), for the RF/antenna side of
+avionics, or when no bus architecture or ICD exists.
+
+## Deliverable contract
+
+The role produces:
+1. **Avionics Data Bus Loading and Protocol Assessment** — per the
+   filled `templates/bus-assessment-template.md`: the bus architecture
+   inventory, ARINC 429 per-line loading (word rate, bus load, %
+   utilization, capacity and design-guideline verdicts, headroom), the
+   MIL-STD-1553 BC minor-frame loading (wire words, schedule time, %
+   utilization, budget headroom), the AFDX VL bandwidth budget (BAG,
+   frame, bandwidth, link utilization, latency and jitter margins), and
+   the received-word protocol conformance checks.
+2. **Findings / observations** — buses over or near the design
+   guidelines, VL sets over the link budget, conformance mismatches.
+3. **Evidence bundle** — `evidence/{model,gates,provenance}.json`
+   (docs/PROTOCOL.md) so any harness can consume the computed numbers.
+
+## Workflow (stages → bound skills)
+
+| Stage | Loads skill | Produces |
+|---|---|---|
+| 1. Architecture inventory | data-bus/arinc429-protocol, data-bus/mil-std-1553 | bus list: TX lines, BC/RT schedule, AFDX VLs |
+| 2. ARINC 429 decode review | data-bus/arinc429-protocol | word decode of received traffic: label, SDI, SSM, parity, BNR/BCD |
+| 3. ARINC 429 bus loading | data-bus/arinc429-bus-loading | per-line total word rate, load bps, % utilization, headroom |
+| 4. MIL-STD-1553 word review | data-bus/mil-std-1553 | command/status/data decode + message classification |
+| 5. MIL-STD-1553 bus loading | data-bus/mil-std-1553-bus-loading | BC minor-frame wire words, schedule time, % utilization |
+| 6. AFDX bandwidth check | data-bus/arinc664-afdx | VL bandwidth from BAG/frame, link budget, latency/jitter margins |
+| 7. Conformance gate | all five leaves | parity/label/classification checks vs ICD expectations |
+| 8. Assessment package | arinc429-bus-loading + arinc664-afdx | filled deliverable + findings + evidence bundle |
+
+## Evidence gates
+
+- Stage 3 done = every ARINC 429 line has a computed utilization % and a
+  verdict against the 80% guideline.
+- Stage 5 done = the BC minor-frame schedule has wire-word counts, total
+  schedule time, and % utilization with budget headroom.
+- Stage 6 done = every AFDX VL has BAG-consistent bandwidth and the VL
+  set fits the 100 Mbps link (or the overrun is a recorded finding).
+- Stage 7 done = received words decode with parity OK and match the ICD
+  label/classification expectation.
+- FINAL = every computed number in the deliverable comes from the bound
+  leaf formulas (36 bit-times/word, 24 us/1553 wire word, frame x 8 / BAG)
+  and is present in the evidence model.
+
+## Boundary / forbidden
+
+- NEVER issue bus architecture sign-off or claim network qualification
+  approval — the program's systems/data-bus sign-off chain signs.
+- NEVER assert a loading margin without the computed budget behind it
+  (the core computes it; the doc quotes it).
+- NEVER reproduce ARINC 429 / ARINC 664 / MIL-STD-1553 standard text —
+  summaries and original structures only.
+- Output is a DRAFT for human review — mark it draft.
+
+## Verification
+
+The role runs STANDALONE: `core/data_bus_avionics_core.py` is an
+executable engine that decodes ARINC 429 and MIL-STD-1553 words,
+computes ARINC 429 bus load %, MIL-STD-1553 bus load %, AFDX VL
+bandwidth, BUILDS the assessment, and gate-checks deliverables. No
+AeroSkills checkout required.
+
+Run the role:
+```bash
+python3 cli.py build --out assessment.md            # example item
+python3 cli.py build --bundle --out assessment.md   # + evidence bundle
+python3 cli.py build --profile <program>.json       # customer tailoring
+python3 cli.py check --file assessment.md           # gate-check
+```
+
+Tests:
+- tests/test_data_bus_avionics_core.py: domain rules with real computed
+  anchors (ARINC 429 word 1611876616 decodes to label 010/SDI 1/data
+  1234; 1553 command word 93316; AFDX VL 4 ms/1518 -> 3.036 Mbps; 30 VLs
+  at 4 ms/1518 = 91.08% of the link) + builder + gates + standalone
+  (no skills repo)
+- tests/test_role_data_bus_avionics_engineer.py: bound-skill resolution
+  (skips if the skills repo is absent) + workflow + template present +
+  boundaries
+- tests/test_bundle_protocol.py: --bundle emits valid evidence; when the
+  skills library is present the five bound leaf logic modules dispatch
+  and agree with the core within 1e-6
+
+Bound skills in Aero Agent Skills deepen individual stages when the
+library is present (word decode, BNR/BCD coding, schedule budgets, AFDX
+sizing); the core engine does not depend on them.
+
+## Compliance
+
+- arinc-429 / arinc-664 TIER-2 reference-only, mil-std-1553 TIER-1
+  public domain, per standards-map; no verbatim anywhere.
+- SOURCES.md records acquisition/extraction/verification status.
+- Author: ashfordeOU (roles wave R1).