Merge pull request #4 from Coding-Autopilot-System/hardening/enterprise-audit-20260610

OgeonX-Ai · web-flow · commit 05cdc3698dc0 · 2026-06-11T13:56:02.000+03:00
feat: harden deterministic CI repair demo
diff --git a/.github/workflows/autopilot-create-issue.yml b/.github/workflows/autopilot-create-issue.yml
@@ -1,11 +1,15 @@
-name: Autopilot Issue Intake
+﻿name: Autopilot Issue Intake
 
 on:
   workflow_run:
     # Update this list in each repo to match the workflows you want monitored.
     workflows: ["Demo CI"]
     types: [completed]
 
+concurrency:
+  group: autopilot-intake-${{ github.event.workflow_run.head_sha }}
+  cancel-in-progress: false
+
 permissions:
   actions: read
   issues: write
@@ -15,9 +19,10 @@ jobs:
   create-issue:
     if: ${{ github.event.workflow_run.conclusion == 'failure' }}
     runs-on: ubuntu-latest
+    timeout-minutes: 5
     steps:
       - name: Create or update issue
-        uses: actions/github-script@v7
+        uses: actions/github-script@60a0d83039c74a4aee543508d2ffcb1c3799cdea # v7.0.1
         with:
           script: |
             const run = context.payload.workflow_run;
@@ -57,6 +62,10 @@ jobs:
               `Branch: ${run.head_branch}`,
               `SHA: ${run.head_sha}`,
               `Run: ${run.html_url}`,
+              `Attempt: ${run.run_attempt}`,
+              `Event: ${run.event}`,
+              `Actor: ${run.actor?.login || 'unknown'}`,
+              `Conclusion: ${run.conclusion}`,
               '',
               'Failed steps:',
               stepSummary,
@@ -92,6 +101,7 @@ jobs:
                 owner,
                 repo,
                 issue_number: issue.number,
+                state: 'open',
                 body,
               });
               await github.rest.issues.addLabels({
diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
@@ -1,29 +1,40 @@
-name: CI
+﻿name: CI
 
 on:
   push:
     branches: [main]
   pull_request:
     branches: [main]
 
+permissions:
+  contents: read
+
 jobs:
   validate:
     runs-on: ubuntu-latest
+    timeout-minutes: 5
     steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+
+      - name: Install validation dependencies
+        run: python -m pip install --requirement requirements-dev.txt
 
       - name: Validate workflow YAML
+        shell: python
         run: |
-          python -c "
-          import yaml, glob
-          files = glob.glob('.github/workflows/*.yml')
-          errors = []
-          for f in files:
-              try:
-                  with open(f) as fh:
-                      yaml.safe_load(fh)
-                  print('  OK:', f)
-              except yaml.YAMLError as e:
-                  print('  WARN:', f, '-', str(e).split('\n')[0])
-          print('Validated', len(files), 'workflow files')
-          "
+          import pathlib
+          import yaml
+
+          files = sorted(pathlib.Path('.github/workflows').glob('*.yml'))
+          if not files:
+              raise SystemExit('No workflow files found')
+
+          for path in files:
+              with path.open(encoding='utf-8') as stream:
+                  yaml.safe_load(stream)
+              print(f'OK: {path}')
+
+          print(f'Validated {len(files)} workflow files')
+
+      - name: Run workflow contract tests
+        run: python -m unittest discover -s tests -v
diff --git a/.github/workflows/demo-ci.yml b/.github/workflows/demo-ci.yml
@@ -1,14 +1,30 @@
-name: Demo CI
+﻿name: Demo CI
 
 on:
   workflow_dispatch:
+    inputs:
+      simulate_failure:
+        description: "Fail the demo job to exercise Autopilot intake"
+        required: true
+        type: boolean
+        default: false
   push:
     branches: [main]
 
+permissions:
+  contents: read
+
 jobs:
   demo:
     runs-on: ubuntu-latest
+    timeout-minutes: 5
     steps:
       - name: Demo sanity
         run: |
           echo "Simulated demo step (non-blocking)."
+
+      - name: Simulate repairable failure
+        if: ${{ inputs.simulate_failure }}
+        run: |
+          echo "Intentional demo failure requested."
+          exit 1
diff --git a/.gitignore b/.gitignore
@@ -0,0 +1,2 @@
+__pycache__/
+*.py[cod]
diff --git a/.planning/phases/01-enterprise-hardening/01-SUMMARY.md b/.planning/phases/01-enterprise-hardening/01-SUMMARY.md
@@ -0,0 +1,7 @@
+# Enterprise Hardening Summary
+
+The GSD audit classified eight findings as auto-fixable and two GitHub administration items as manual-only. All eight repository findings were fixed, tested, and recorded in `memory/examples/`.
+
+The demo now provides a deliberate failure switch, strict CI validation, immutable Action references, least-privilege workflow boundaries, executable workflow contracts, serialized intake, richer audit evidence, and a repeatable operator runbook.
+
+F-08 failed its first contract test because the intake update path did not reopen closed issues. That attempt was reverted, corrected, and verified on retry.
diff --git a/.planning/phases/01-enterprise-hardening/01-UAT.md b/.planning/phases/01-enterprise-hardening/01-UAT.md
@@ -0,0 +1,23 @@
+﻿# Enterprise Hardening UAT
+
+**Date:** 2026-06-10
+**Source:** `gsd-audit-fix --severity all --max 8`
+**Scope:** demo reproducibility, security, tests, CI, user journey, failure modes, and docs
+
+## Classification
+
+| ID | Severity | Classification | Finding | Files |
+|---|---|---|---|---|
+| F-01 | high | auto-fixable | The documented demo cannot produce a failure because Demo CI always succeeds. | `.github/workflows/demo-ci.yml`, `README.md` |
+| F-02 | high | auto-fixable | CI prints YAML parse warnings but still exits successfully. | `.github/workflows/ci.yml` |
+| F-03 | high | auto-fixable | Third-party Actions use mutable major-version tags instead of immutable commit SHAs. | `.github/workflows/ci.yml`, `.github/workflows/autopilot-create-issue.yml` |
+| F-04 | medium | auto-fixable | CI and Demo CI do not declare least-privilege token permissions or job timeouts. | `.github/workflows/ci.yml`, `.github/workflows/demo-ci.yml` |
+| F-05 | medium | auto-fixable | No automated contract tests protect workflow behavior and security invariants. | `tests/test_workflows.py`, `.github/workflows/ci.yml` |
+| F-06 | medium | auto-fixable | Intake processing has no concurrency guard or timeout, allowing duplicate/racing issue updates. | `.github/workflows/autopilot-create-issue.yml` |
+| F-07 | low | auto-fixable | Intake issues omit run attempt, event, actor, and failure conclusion from the audit evidence. | `.github/workflows/autopilot-create-issue.yml` |
+| F-08 | low | auto-fixable | The demo runbook lacks prerequisites, reset steps, and failure-mode troubleshooting; repeat demonstrations can leave a matching intake issue closed. | `.github/workflows/autopilot-create-issue.yml`, `README.md` |
+
+## Manual-only Findings
+
+- M-01: Protect `main` and require the `CI` status check. The GitHub API reports that `main` is not protected.
+- M-02: Configure the organization Actions policy to allow only approved actions and require immutable SHA pinning where supported. The current GitHub token cannot read or change this setting.
diff --git a/.planning/phases/01-enterprise-hardening/01-VERIFICATION.md b/.planning/phases/01-enterprise-hardening/01-VERIFICATION.md
@@ -0,0 +1,33 @@
+# Enterprise Hardening Verification
+
+**Date:** 2026-06-10
+**Branch:** `hardening/enterprise-audit-20260610`
+**Result:** passed with manual GitHub administration gaps
+
+## Finding Status
+
+| ID | Result | Evidence |
+|---|---|---|
+| F-01 | passed | `simulate_failure` is explicit and defaults to false |
+| F-02 | passed | invalid YAML now fails CI; dependency is pinned |
+| F-03 | passed | third-party Actions use verified 40-character commit SHAs |
+| F-04 | passed | routine jobs declare read-only permissions and five-minute timeouts |
+| F-05 | passed | seven workflow contract tests run in CI; bytecode is ignored |
+| F-06 | passed | intake updates are serialized by head SHA and time-bounded |
+| F-07 | passed | issues record attempt, event, actor, and conclusion |
+| F-08 | passed after one reverted attempt | matching closed issues reopen; runbook covers prerequisites, reset, and troubleshooting |
+
+## Verification Commands
+
+- `python -m unittest discover -s tests -v`: 7 passed
+- strict PyYAML parse of `.github/workflows/*.yml`: 3 validated
+- PowerShell parser check of `scripts/record-fix.ps1`: passed
+- `git diff --check origin/main...HEAD`: passed
+- clean worktree after tests: passed
+
+## Manual Gaps
+
+- Protect `main` and require the `CI` status check.
+- Configure the organization Actions allow-list and immutable pinning policy.
+- Run the live failure-to-issue-to-repair journey after this branch is merged because `workflow_dispatch` uses the default-branch workflow definition.
+- Install `actionlint` in the workstation bootstrap and add it to CI; it was not available for this verification.
diff --git a/README.md b/README.md
@@ -1,4 +1,4 @@
-# autopilot-demo
+﻿# autopilot-demo
 
 [![CI](https://github.com/Coding-Autopilot-System/autopilot-demo/actions/workflows/ci.yml/badge.svg?branch=main)](https://github.com/Coding-Autopilot-System/autopilot-demo/actions/workflows/ci.yml)
 [![Demo CI](https://github.com/Coding-Autopilot-System/autopilot-demo/actions/workflows/demo-ci.yml/badge.svg?branch=main)](https://github.com/Coding-Autopilot-System/autopilot-demo/actions/workflows/demo-ci.yml)
@@ -47,7 +47,7 @@ gh pr list -R Coding-Autopilot-System/autopilot-demo
 
 ## Demo runbook
 
-1. Trigger [`.github/workflows/demo-ci.yml`](.github/workflows/demo-ci.yml) to produce a known failure signal.
+1. Trigger [`.github/workflows/demo-ci.yml`](.github/workflows/demo-ci.yml) with `simulate_failure=true` to produce a known failure signal. Pushes and default dispatches remain green.
 2. Confirm [`.github/workflows/autopilot-create-issue.yml`](.github/workflows/autopilot-create-issue.yml) creates an `autofix + queued` issue.
 3. Watch `autopilot-core` pick up the issue and open a PR back into this repo.
 4. Use this repo's issue, branch, and PR history as the audit trail for the demo.
@@ -72,3 +72,16 @@ gh pr list -R Coding-Autopilot-System/autopilot-demo
 - [autopilot-core](https://github.com/Coding-Autopilot-System/autopilot-core) - operator control plane
 - [ci-autopilot](https://github.com/Coding-Autopilot-System/ci-autopilot) - worker/runtime reference
 - [Coding-Autopilot-System org](https://github.com/Coding-Autopilot-System)
+## Prerequisites and expected result
+
+- Authenticate GitHub CLI with `gh auth status` and confirm Actions are enabled for this repository.
+- Run the `autopilot-core` operator with access to this repository before triggering the failure.
+- Expect Demo CI to fail, Autopilot Issue Intake to create or reopen one `autofix + queued` issue, and the operator to propose a pull request.
+
+## Reset and troubleshooting
+
+1. Close the completed intake issue and merge or close its repair pull request before the next demonstration.
+2. Re-run with `simulate_failure=true`; intake reopens the matching issue when the same commit is demonstrated again.
+3. If no issue appears, inspect `gh run list -R Coding-Autopilot-System/autopilot-demo --workflow autopilot-create-issue.yml` and confirm the failed run was named `Demo CI`.
+4. If the issue remains queued, verify the `autopilot-core` operator is running and can read issues and create branches and pull requests in this repository.
+5. If CI fails before the demo step, run `python -m unittest discover -s tests -v` locally and repair the workflow contract violation first.
diff --git a/memory/examples/20260610_f01_reproducible-failure.md b/memory/examples/20260610_f01_reproducible-failure.md
@@ -0,0 +1,8 @@
+﻿# F-01 reproducible demo failure
+
+Issue Description: The documented demo could not produce a failure because Demo CI always succeeded.
+State: Pushes and manual dispatches both ran only a non-blocking step.
+Action: Added an explicit `simulate_failure` dispatch input and documented the required command.
+Result: Operators can intentionally exercise intake while normal pushes and default dispatches stay green.
+Diff Patch: Updated `.github/workflows/demo-ci.yml` and `README.md`.
+Rationale: A demo repair pipeline must expose a safe, deliberate, and reproducible failure path.
diff --git a/memory/examples/20260610_f02_strict-yaml-validation.md b/memory/examples/20260610_f02_strict-yaml-validation.md
@@ -0,0 +1,8 @@
+﻿# F-02 strict YAML validation
+
+Issue Description: CI printed YAML parsing warnings but always returned success.
+State: Invalid workflow YAML could pass the portfolio CI check.
+Action: Made parsing errors fatal and pinned the YAML parser dependency.
+Result: CI now fails closed when workflow YAML is missing or invalid.
+Diff Patch: Updated `.github/workflows/ci.yml` and added `requirements-dev.txt`.
+Rationale: Validation jobs must fail on the condition they claim to validate.
diff --git a/memory/examples/20260610_f03_immutable-actions.md b/memory/examples/20260610_f03_immutable-actions.md
@@ -0,0 +1,8 @@
+﻿# F-03 immutable Action references
+
+Issue Description: Workflows trusted mutable major-version Action tags.
+State: Upstream tag movement could change executed code without a repository diff.
+Action: Pinned checkout v4.2.2 and github-script v7.0.1 to verified commit SHAs.
+Result: Third-party workflow code is immutable and remains human-readable through version comments.
+Diff Patch: Updated CI and intake workflow Action references.
+Rationale: Immutable references reduce CI supply-chain risk.
diff --git a/memory/examples/20260610_f04_workflow-boundaries.md b/memory/examples/20260610_f04_workflow-boundaries.md
@@ -0,0 +1,8 @@
+﻿# F-04 least-privilege workflow boundaries
+
+Issue Description: CI and Demo CI relied on repository-default token permissions and had no execution timeout.
+State: Jobs could inherit broader permissions and run without a repository-defined time bound.
+Action: Declared read-only contents permission and five-minute job timeouts.
+Result: Routine workflows now have explicit least privilege and bounded execution.
+Diff Patch: Updated `.github/workflows/ci.yml` and `.github/workflows/demo-ci.yml`.
+Rationale: Security boundaries should be visible and enforced in source control.
diff --git a/memory/examples/20260610_f05_workflow-contract-tests.md b/memory/examples/20260610_f05_workflow-contract-tests.md
@@ -0,0 +1,8 @@
+﻿# F-05 workflow contract tests
+
+Issue Description: No automated tests protected workflow behavior and security invariants.
+State: Regressions in YAML validity, Action pinning, permissions, timeouts, or demo failure behavior depended on review alone.
+Action: Added standard-library workflow contract tests and executed them from CI.
+Result: Core demo and security invariants now fail visibly in pull requests.
+Diff Patch: Added `tests/test_workflows.py` and updated `.github/workflows/ci.yml`.
+Rationale: Workflow code needs executable contracts like application code.
diff --git a/memory/examples/20260610_f06_serialized-intake.md b/memory/examples/20260610_f06_serialized-intake.md
@@ -0,0 +1,8 @@
+﻿# F-06 serialized intake processing
+
+Issue Description: Intake processing had no concurrency guard or execution timeout.
+State: Multiple failures for the same commit could race while updating the same signature issue.
+Action: Serialized intake by workflow head SHA, preserved queued runs, and bounded the job to five minutes.
+Result: Intake updates are deterministic and cannot run indefinitely.
+Diff Patch: Updated the intake workflow and its contract tests.
+Rationale: Idempotent issue handling still needs race protection at the workflow boundary.
diff --git a/memory/examples/20260610_f07_intake-audit-context.md b/memory/examples/20260610_f07_intake-audit-context.md
@@ -0,0 +1,8 @@
+﻿# F-07 richer intake audit evidence
+
+Issue Description: Intake issues omitted operational context needed to reconstruct a failure.
+State: Issues recorded workflow, branch, SHA, run URL, and failed steps only.
+Action: Added run attempt, triggering event, actor, and conclusion to every intake issue body.
+Result: Each issue now carries a stronger self-contained audit trail.
+Diff Patch: Updated the intake workflow and its contract tests.
+Rationale: Operational evidence should support diagnosis without requiring immediate API access.
diff --git a/memory/examples/20260610_f08_repeatable-runbook.md b/memory/examples/20260610_f08_repeatable-runbook.md
@@ -0,0 +1,8 @@
+﻿# F-08 repeatable operator runbook
+
+Issue Description: The runbook lacked prerequisites, reset steps, and troubleshooting, while repeat demonstrations could leave the matching issue closed.
+State: Operators had no defined recovery path and intake updated closed matching issues without reopening them.
+Action: Documented prerequisites, expected results, reset and troubleshooting; reopen matching intake issues on repeat failures.
+Result: The user journey is repeatable and common failure modes have concrete diagnostics.
+Diff Patch: Updated the intake workflow, README, UAT record, and contract tests.
+Rationale: A portfolio demo must be operable repeatedly by someone other than its author.
diff --git a/requirements-dev.txt b/requirements-dev.txt
@@ -0,0 +1 @@
+PyYAML==6.0.3
diff --git a/tests/test_workflows.py b/tests/test_workflows.py
@@ -0,0 +1,59 @@
+﻿import pathlib
+import re
+import unittest
+
+import yaml
+
+
+ROOT = pathlib.Path(__file__).resolve().parents[1]
+WORKFLOWS = ROOT / ".github" / "workflows"
+
+
+class WorkflowContractTests(unittest.TestCase):
+    def workflow_text(self, name: str) -> str:
+        return (WORKFLOWS / name).read_text(encoding="utf-8-sig")
+
+    def test_all_workflows_are_valid_yaml(self) -> None:
+        paths = sorted(WORKFLOWS.glob("*.yml"))
+        self.assertTrue(paths, "expected at least one workflow")
+        for path in paths:
+            with self.subTest(path=path.name):
+                self.assertIsInstance(yaml.safe_load(path.read_text(encoding="utf-8-sig")), dict)
+
+    def test_third_party_actions_are_pinned_to_commits(self) -> None:
+        action_ref = re.compile(r"uses:\s+[^./\s][^@\s]*@([^\s#]+)")
+        for path in WORKFLOWS.glob("*.yml"):
+            for ref in action_ref.findall(path.read_text(encoding="utf-8-sig")):
+                with self.subTest(path=path.name, ref=ref):
+                    self.assertRegex(ref, r"^[0-9a-f]{40}$")
+
+    def test_routine_workflows_are_read_only_and_bounded(self) -> None:
+        for name in ("ci.yml", "demo-ci.yml"):
+            text = self.workflow_text(name)
+            with self.subTest(name=name):
+                self.assertIn("permissions:\n  contents: read", text.replace("\r\n", "\n"))
+                self.assertIn("timeout-minutes: 5", text)
+
+    def test_intake_is_serialized_and_bounded(self) -> None:
+        text = self.workflow_text("autopilot-create-issue.yml")
+        self.assertIn("group: autopilot-intake-${{ github.event.workflow_run.head_sha }}", text)
+        self.assertIn("cancel-in-progress: false", text)
+        self.assertIn("timeout-minutes: 5", text)
+    def test_intake_records_audit_context(self) -> None:
+        text = self.workflow_text("autopilot-create-issue.yml")
+        for field in ("Attempt:", "Event:", "Actor:", "Conclusion:"):
+            with self.subTest(field=field):
+                self.assertIn(field, text)
+    def test_existing_intake_issue_is_reopened(self) -> None:
+        text = self.workflow_text("autopilot-create-issue.yml")
+        self.assertIn("state: 'open'", text)
+    def test_demo_failure_is_explicitly_opt_in(self) -> None:
+        text = self.workflow_text("demo-ci.yml")
+        self.assertIn("simulate_failure:", text)
+        self.assertIn("default: false", text)
+        self.assertIn("if: ${{ inputs.simulate_failure }}", text)
+        self.assertIn("exit 1", text)
+
+
+if __name__ == "__main__":
+    unittest.main()
