Governance: AI-contribution rules, AI statement check, pinned actions, workflow lint, dependency review

.github/CODEOWNERS

@@ -1,14 +1,26 @@
 # Pull requests that touch these files need a maintainer's approval (branch
 # protection: "Require review from Code Owners"). They hold recorded results,
 # the tests and the settings that decide which tests run, and the CI that runs
-# them; a pull request must not edit them to make itself pass.
-/test/            @Hananel-Hazan
-/.github/         @Hananel-Hazan
-/pyproject.toml   @Hananel-Hazan
-/poetry.lock      @Hananel-Hazan
-/pytest.ini       @Hananel-Hazan
-/.pytest.ini      @Hananel-Hazan
-/tox.ini          @Hananel-Hazan
-/setup.cfg        @Hananel-Hazan
-/.coveragerc      @Hananel-Hazan
-/conftest.py      @Hananel-Hazan
+# them; a pull request must not edit them to make itself pass. The rest run
+# automatically (editor, container and build configuration), decide what gets
+# installed, or define the learning rules the tests pin to their papers.
+/test/                     @Hananel-Hazan
+/.github/                  @Hananel-Hazan
+/pyproject.toml            @Hananel-Hazan
+/poetry.lock               @Hananel-Hazan
+/pytest.ini                @Hananel-Hazan
+/.pytest.ini               @Hananel-Hazan
+/tox.ini                   @Hananel-Hazan
+/setup.cfg                 @Hananel-Hazan
+/.coveragerc               @Hananel-Hazan
+conftest.py                @Hananel-Hazan
+/.vscode/                  @Hananel-Hazan
+/.devcontainer/            @Hananel-Hazan
+/.pre-commit-config.yaml   @Hananel-Hazan
+/Dockerfile                @Hananel-Hazan
+/.dockerignore             @Hananel-Hazan
+/.readthedocs.yaml         @Hananel-Hazan
+/docs/source/conf.py       @Hananel-Hazan
+/bindsnet/learning/        @Hananel-Hazan
+/SECURITY.md               @Hananel-Hazan
+/CONTRIBUTING.md           @Hananel-Hazan

.github/PULL_REQUEST_TEMPLATE.md

@@ -23,7 +23,15 @@
       - `test/api_surface.json` (public API)
       - `FIXED_BUGS` in `test/network/test_all_classes.py`
 - [ ] I added an entry to the `Unreleased` section of `CHANGELOG.md`.
-- [ ] If I used an AI tool to write or find this change, I name it here:
+
+## AI assistance
+
+<!-- Tick exactly one box (checked by the `AI statement` check). See "AI-assisted contributions" in CONTRIBUTING.md. -->
+
+- [ ] No AI tools were used.
+- [ ] AI tools were used, and I (a human) have read every changed line, run the tests myself, and can explain the change in my own words.
+
+Tools used and for which parts:
 
 ## Why the results or the API changed (if they did)
 

.github/workflows/ai-statement.yml

@@ -0,0 +1,42 @@
+# Checks that a pull request fills in the "AI assistance" section of the template
+# (CONTRIBUTING.md, "AI-assisted contributions"). It reads the description only and
+# runs no code from the pull request.
+name: AI statement
+
+on:
+  pull_request:
+    branches: [ master ]
+    types: [ opened, edited, synchronize, reopened ]
+
+permissions:
+  contents: read
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: true
+
+jobs:
+  ai-statement:
+    if: github.event.pull_request.user.login != 'dependabot[bot]'
+    runs-on: ubuntu-latest
+    steps:
+      # The description is passed as an environment variable, never pasted into the
+      # script, so text in it cannot run as code.
+      - name: Check the AI assistance section
+        env:
+          PR_BODY: ${{ github.event.pull_request.body }}
+        shell: python
+        run: |
+          import os, re, sys
+
+          body = os.environ.get("PR_BODY") or ""
+          none = re.search(r"^- \[[xX]\] No AI tools were used", body, re.M)
+          used = re.search(r"^- \[[xX]\] AI tools were used, and I \(a human\) have read every changed line", body, re.M)
+          tools = re.search(r"^Tools used and for which parts:[ \t]*(\S.*)?$\n?((?:(?!^#).*\n?)*)", body, re.M)
+          described = bool(tools and ((tools.group(1) or "") + tools.group(2)).strip())
+          if bool(none) == bool(used):
+              sys.exit("Tick exactly one box in the 'AI assistance' section of the pull request description "
+                       "(see CONTRIBUTING.md, 'AI-assisted contributions').")
+          if used and not described:
+              sys.exit("Say which AI tools were used and for which parts, under 'Tools used and for which parts:'.")
+          print("AI assistance section filled in.")

.github/workflows/bench.yml

@@ -26,12 +26,13 @@ jobs:
     runs-on: ubuntu-latest
     timeout-minutes: 20
     steps:
-    - uses: actions/checkout@v7
+    - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
       with:
         fetch-depth: 0  # the base commit must be present for git worktree
+        persist-credentials: false
     - name: Install Poetry
       run: pipx install "poetry==${POETRY_VERSION}"
-    - uses: actions/setup-python@v7
+    - uses: actions/setup-python@6d4fd22b20465ccc0f7354476588e8df7c9ca0e0 # v7.1.0
       with:
         python-version: "3.13"
         cache: poetry
@@ -49,7 +50,7 @@ jobs:
         poetry run python .github/scripts/bench_ab.py
         --base "$(git rev-parse HEAD^1)" --head "$HEAD_SHA" --fail-above 2
         --out bench-out --summary "$GITHUB_STEP_SUMMARY"
-    - uses: actions/upload-artifact@v7
+    - uses: actions/upload-artifact@cf430e030ddbb5b0abf93d22962f4752f3646cd9 # v7.0.2
       if: always()
       with:
         name: bench-ab

.github/workflows/black.yml

@@ -18,8 +18,10 @@ jobs:
   lint:
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@v7
-      - uses: actions/setup-python@v7
+      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
+        with:
+          persist-credentials: false
+      - uses: actions/setup-python@6d4fd22b20465ccc0f7354476588e8df7c9ca0e0 # v7.1.0
         with:
           python-version: "3.13"
       # Use the black and isort versions pinned in poetry.lock, so CI and local

.github/workflows/dependency-review.yml

@@ -0,0 +1,25 @@
+# Lists the packages a pull request adds or changes (pyproject.toml, poetry.lock,
+# the Dockerfile's base image, GitHub Actions) and fails if any has a known
+# vulnerability of moderate severity or worse. It reads GitHub's dependency graph
+# only and runs no code from the pull request.
+name: Dependency review
+
+on:
+  pull_request:
+    branches: [ master ]
+
+permissions:
+  contents: read
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: true
+
+jobs:
+  dependency-review:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/dependency-review-action@a1d282b36b6f3519aa1f3fc636f609c47dddb294 # v5.0.0
+        with:
+          fail-on-severity: moderate
+          comment-summary-in-pr: never

.github/workflows/docker.yml

@@ -22,7 +22,9 @@ jobs:
   docker:
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@v7
+      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
+        with:
+          persist-credentials: false
       # The image is about 7 GB (torch bundles its CUDA libraries); make room.
       - name: Free runner disk space
         run: sudo rm -rf /usr/share/dotnet /usr/local/lib/android /opt/ghc /opt/hostedtoolcache/CodeQL

.github/workflows/publish.yml

@@ -12,8 +12,10 @@ jobs:
   build:
     runs-on: ubuntu-latest
     steps:
-    - uses: actions/checkout@v7
-    - uses: actions/setup-python@v7
+    - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
+      with:
+        persist-credentials: false
+    - uses: actions/setup-python@6d4fd22b20465ccc0f7354476588e8df7c9ca0e0 # v7.1.0
       with:
         python-version: "3.13"
     - name: Build sdist and wheel
@@ -30,7 +32,7 @@ jobs:
           echo "Release tag $GITHUB_REF_NAME does not match pyproject version $version"
           exit 1
         fi
-    - uses: actions/upload-artifact@v7
+    - uses: actions/upload-artifact@cf430e030ddbb5b0abf93d22962f4752f3646cd9 # v7.0.2
       with:
         name: dist
         path: dist/
@@ -44,8 +46,8 @@ jobs:
     permissions:
       id-token: write
     steps:
-    - uses: actions/download-artifact@v8
+    - uses: actions/download-artifact@9000827ccba6bdab643e8b6fd33ac0654aef8333 # v8.0.2
       with:
         name: dist
         path: dist/
-    - uses: pypa/gh-action-pypi-publish@release/v1
+    - uses: pypa/gh-action-pypi-publish@7245906fb61fa33a13554f9d5fb856fe61758086 # v1.14.3

.github/workflows/python-app.yml

@@ -23,10 +23,12 @@ jobs:
     name: build
     runs-on: ubuntu-latest
     steps:
-    - uses: actions/checkout@v7
+    - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
+      with:
+        persist-credentials: false
     - name: Install Poetry
       run: pipx install "poetry==${POETRY_VERSION}"
-    - uses: actions/setup-python@v7
+    - uses: actions/setup-python@6d4fd22b20465ccc0f7354476588e8df7c9ca0e0 # v7.1.0
       with:
         python-version: "3.13"
         cache: poetry
@@ -50,7 +52,7 @@ jobs:
     # below the value recorded in the script (see its docstring).
     - name: Check coverage floor
       run: python .github/scripts/coverage_floor.py coverage.xml
-    - uses: actions/upload-artifact@v7
+    - uses: actions/upload-artifact@cf430e030ddbb5b0abf93d22962f4752f3646cd9 # v7.0.2
       with:
         name: coverage-xml
         path: coverage.xml
@@ -63,10 +65,12 @@ jobs:
       matrix:
         python-version: ["3.11", "3.12"]
     steps:
-    - uses: actions/checkout@v7
+    - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
+      with:
+        persist-credentials: false
     - name: Install Poetry
       run: pipx install "poetry==${POETRY_VERSION}"
-    - uses: actions/setup-python@v7
+    - uses: actions/setup-python@6d4fd22b20465ccc0f7354476588e8df7c9ca0e0 # v7.1.0
       with:
         python-version: ${{ matrix.python-version }}
         cache: poetry

.github/workflows/test-guard.yml

@@ -33,12 +33,13 @@ jobs:
     runs-on: ubuntu-latest
     timeout-minutes: 15
     steps:
-    - uses: actions/checkout@v7
+    - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
       with:
         fetch-depth: 2  # HEAD is the test merge; HEAD^1 is master it merges into
+        persist-credentials: false
     - name: Install Poetry
       run: pipx install "poetry==${POETRY_VERSION}"
-    - uses: actions/setup-python@v7
+    - uses: actions/setup-python@6d4fd22b20465ccc0f7354476588e8df7c9ca0e0 # v7.1.0
       with:
         python-version: "3.13"
         cache: poetry
@@ -57,12 +58,13 @@ jobs:
     runs-on: ubuntu-latest
     timeout-minutes: 20
     steps:
-    - uses: actions/checkout@v7
+    - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
       with:
         fetch-depth: 2
+        persist-credentials: false
     - name: Install Poetry
       run: pipx install "poetry==${POETRY_VERSION}"
-    - uses: actions/setup-python@v7
+    - uses: actions/setup-python@6d4fd22b20465ccc0f7354476588e8df7c9ca0e0 # v7.1.0
       with:
         python-version: "3.13"
         cache: poetry

.github/workflows/workflow-lint.yml

@@ -0,0 +1,36 @@
+# Audits the workflow files themselves: zizmor looks for unsafe patterns (untrusted
+# text reaching a shell, unpinned actions, leaked credentials), actionlint for mistakes.
+# Both run from version-pinned packages, not third-party actions.
+name: Workflow lint
+
+on:
+  push:
+    branches: [ master ]
+    paths: [ ".github/workflows/**" ]
+  pull_request:
+    branches: [ master ]
+    paths: [ ".github/workflows/**" ]
+
+permissions:
+  contents: read
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: true
+
+jobs:
+  workflow-lint:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
+        with:
+          persist-credentials: false
+      - uses: actions/setup-python@6d4fd22b20465ccc0f7354476588e8df7c9ca0e0 # v7.1.0
+        with:
+          python-version: "3.13"
+      - name: zizmor
+        run: pipx run "zizmor==1.30.1" --offline .github/workflows
+      - name: actionlint
+        run: |
+          go install github.com/rhysd/actionlint/cmd/actionlint@v1.7.12
+          "$(go env GOPATH)/bin/actionlint"

CHANGELOG.md

@@ -6,6 +6,30 @@ see the [GitHub releases / tags](https://github.com/BindsNET/bindsnet/releases).
 
 ## [Unreleased]
 
+### Added
+- Contributor rules for AI-assisted work (`CONTRIBUTING.md`, "AI-assisted
+  contributions"): AI tools are welcome, a human must have read and run every change,
+  and autonomous agents may not submit. The pull request template has an "AI
+  assistance" section, and the `AI statement` workflow checks that it is filled in.
+- `Workflow lint` workflow: zizmor 1.30.1 and actionlint 1.7.12 audit changes to
+  `.github/workflows/`.
+- `Dependency review` workflow: fails a pull request that adds or updates a package
+  with a known vulnerability of moderate severity or worse.
+
+### Changed
+- CI actions are pinned to exact commits instead of version tags, and checkouts no
+  longer keep the GitHub token on disk.
+- `SECURITY.md` now describes the repository controls as they are.
+- `CODEOWNERS` also covers files that run automatically or define tested behaviour:
+  `.vscode/`, `.devcontainer/`, `.pre-commit-config.yaml`, `Dockerfile`,
+  `.dockerignore`, `.readthedocs.yaml`, `docs/source/conf.py`, every `conftest.py`,
+  `bindsnet/learning/`, `SECURITY.md` and `CONTRIBUTING.md`.
+
+### Fixed
+- The CUDA `test_transfer` tests in `test/network/test_nodes.py` and
+  `test/network/test_connections.py` are reported as skipped on machines without a
+  GPU; before, they returned early and were reported as passed.
+
 ## [0.3.5] - 2026-10-29
 
 ### Results changes (read this before upgrading)

CONTRIBUTING.md

@@ -1,5 +1,31 @@
 # Contributing Protocol
 
+## AI-assisted contributions
+
+You are welcome to use AI tools to find bugs in BindsNET and to improve it. A
+person must stay responsible for the work:
+
+- **You are the author.** The person who opens the pull request answers for every
+  line in it, whoever or whatever wrote it.
+- **Read it and run it yourself.** Before you open the pull request, read every
+  changed line, run the tests on your own machine, and make sure you can explain
+  the change in your own words.
+- **No autonomous agents.** An AI agent may not open pull requests, push commits or
+  post comments on its own. A pull request stating that no human has reviewed the
+  code is put on hold until its author has.
+- **Say what you used.** Fill in the "AI assistance" section of the pull request
+  template: which tools, and for which parts. A CI check makes sure the section is
+  filled in.
+- **Answer review questions yourself.** The maintainer may ask you about the design
+  of the change. Please answer in your own words, not with AI-generated text
+  (translation and grammar help are fine).
+
+This follows the policies of NumPy
+(https://numpy.org/devdocs/dev/ai_policy.html) and LLVM
+(https://llvm.org/docs/AIToolPolicy.html).
+
+## Getting started
+
 To clone this project locally, issue
 
 ```shell

SECURITY.md

@@ -57,7 +57,14 @@ BindsNET is a research library, and its git history is part of what users rely
 on. The following controls are in place on this repository:
 
 - Force-pushes and branch deletions are blocked on every branch.
-- `master` requires a pull request with an approving review.
+- Changes reach `master` through pull requests with signed commits. Each needs
+  an approving review that covers its latest push, the code owners' review for
+  the paths in `.github/CODEOWNERS`, and passing test, formatting, test-guard
+  and speed checks. The maintainer reviews every pull request from an outside
+  contributor before it is merged.
+- Workflows get a read-only token by default, every outside contributor's
+  pull request waits for the maintainer's approval before CI runs, and the
+  actions CI uses are pinned to exact commits.
 
 A supply-chain incident affecting this repository was reported and remediated in
 September 2026; see issue #781 for the full account. No PyPI release was