igraphdata/.github/workflows/format-suggest.yaml at main · igraph/igraphdata · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
# Workflow derived from https://github.com/posit-dev/setup-air/tree/main/examples

on:
  # We keep `pull_request_target` (rather than `pull_request`) so that this job
  # can post review suggestions on pull requests that come from forks, which is
  # our main use case. A fork's `pull_request` run only gets a read-only token,
  # so it cannot comment; `pull_request_target` runs with the base repository's
  # token and can be granted `pull-requests: write`.
  #
  # SECURITY -- this is a classic "pwn request" surface, so the fork's code is
  # treated strictly as *data to be formatted*, never as code to run:
  #
  #   1. The formatter tooling (the `style` action and the air / clang-format
  #      binaries) is never loaded from the PR checkout. This is the critical
  #      point. The naive pattern
  #        - uses: actions/checkout@v6      # ref: fork head
  #        - uses: ./.github/workflows/style
  #      resolves `./...style` from the *checked-out fork*, so an attacker only
  #      has to edit their fork's `style/action.yml` to run arbitrary commands
  #      with our token and secrets. A `owner/repo/path@ref` reference is
  #      fetched from that repository instead of from the workspace, so the
  #      fork controls only the input files, not the code that executes.
  #   2. air and clang-format merely parse and re-print source; they do not
  #      evaluate it. No build / install / test step ever runs the fork's code.
  #   3. The PR is checked out with `persist-credentials: false`, so the token
  #      is not written to disk alongside untrusted files.
  #   4. Permissions are reduced to `pull-requests: write` only, and
  #      `harden-runner` observes (and can block) network egress.
  #
  # https://securitylab.github.com/resources/github-actions-preventing-pwn-requests/
  # https://gh.io/securely-using-pull_request_target
  pull_request_target:

name: format-suggest.yaml

# Deny everything by default; the job below opts back into the single scope it
# needs. This caps the blast radius of the elevated `pull_request_target` token.
permissions: {}

jobs:
  format-suggest:
    name: format-suggest
    runs-on: ubuntu-26.04
    timeout-minutes: 15 # longest run seen: 1 min
    # Only run this job if changes come from a fork.
    # We commit changes directly on the main repository.
    if: github.event.pull_request.head.repo.full_name != github.repository

    permissions:
      # Required to push suggestion comments to the PR
      pull-requests: write

    steps:
      # Defense in depth: block all outbound network traffic except the
      # endpoints the formatters legitimately need. Even if a formatter ever
      # mishandled attacker-controlled input, it could not exfiltrate the token
      # or fetch a second-stage payload.
      #
      # The GitHub endpoints below were confirmed from a harden-runner `audit`
      # run of this workflow (git checkout, setup-air / reviewdog binary
      # downloads, reviewdog posting via the API). The apt.llvm.org / Ubuntu
      # archive endpoints are only hit when a package ships a `.clang-format`
      # file and the `style` action installs clang-format; they are included so
      # C++ consumers of this template are not broken, even though a pure-R
      # package never contacts them.
      #
      # If a run is ever blocked, harden-runner reports the denied endpoint in
      # the job's "Harden Runner" step summary (and the StepSecurity insights
      # for the run) -- add it here and re-run. See the note in the PR for the
      # audit -> block workflow.
      - name: Harden runner
        uses: step-security/harden-runner@bf7454d06d71f1098171f2acdf0cd4708d7b5920 # v2
        with:
          egress-policy: block
          allowed-endpoints: >
            github.com:443
            api.github.com:443
            release-assets.githubusercontent.com:443
            objects.githubusercontent.com:443
            codeload.github.com:443
            apt.llvm.org:443
            azure.archive.ubuntu.com:80

      # The untrusted PR code, checked out at the workspace root. It is DATA
      # only -- nothing below executes it. `allow-unsafe-pr-checkout` is required
      # by actions/checkout@v6 for a fork ref under pull_request_target; it is
      # safe here specifically because the code that runs is fetched from
      # `cynkra/cynkratemplate`, never from this checkout. See the notes above.
      - name: Check out PR code (treated as data)
        uses: actions/checkout@v6
        with:
          ref: ${{ github.event.pull_request.head.sha }}
          persist-credentials: false
          allow-unsafe-pr-checkout: true

      # Runs the trusted `style` action against the PR code at the workspace
      # root. Because the action is fetched from `cynkra/cynkratemplate` rather
      # than resolved out of the workspace, the fork controls only the files
      # being formatted, not the code that runs.
      - name: Format
        uses: cynkra/cynkratemplate/.github/actions/style@main

      - name: Suggest
        uses: reviewdog/action-suggester@2558ba17e65a9039e73764a73009fc05fef28a46 # v1
        with:
          level: error
          fail_level: error
          tool_name: air-and-clang-format