securelint 1.0.0__tar.gz

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (96) hide show
  1. securelint-1.0.0/.gitattributes +3 -0
  2. securelint-1.0.0/.github/workflows/ci.yml +112 -0
  3. securelint-1.0.0/.github/workflows/release.yml +116 -0
  4. securelint-1.0.0/.gitignore +16 -0
  5. securelint-1.0.0/CHANGELOG.md +248 -0
  6. securelint-1.0.0/LICENSE +21 -0
  7. securelint-1.0.0/PKG-INFO +701 -0
  8. securelint-1.0.0/README.md +664 -0
  9. securelint-1.0.0/SECURITY.md +33 -0
  10. securelint-1.0.0/action.yml +208 -0
  11. securelint-1.0.0/benchmarks/compare_bandit.py +237 -0
  12. securelint-1.0.0/benchmarks/real_projects.py +68 -0
  13. securelint-1.0.0/benchmarks/scan_speed.py +71 -0
  14. securelint-1.0.0/docs/real-projects.md +94 -0
  15. securelint-1.0.0/docs/releasing.md +78 -0
  16. securelint-1.0.0/docs/threat-model.md +179 -0
  17. securelint-1.0.0/pyproject.toml +104 -0
  18. securelint-1.0.0/src/securelint/__init__.py +3 -0
  19. securelint-1.0.0/src/securelint/__main__.py +5 -0
  20. securelint-1.0.0/src/securelint/baseline.py +139 -0
  21. securelint-1.0.0/src/securelint/cache.py +103 -0
  22. securelint-1.0.0/src/securelint/changes.py +250 -0
  23. securelint-1.0.0/src/securelint/cli.py +434 -0
  24. securelint-1.0.0/src/securelint/config.py +304 -0
  25. securelint-1.0.0/src/securelint/context.py +127 -0
  26. securelint-1.0.0/src/securelint/dataflow.py +242 -0
  27. securelint-1.0.0/src/securelint/discovery.py +81 -0
  28. securelint-1.0.0/src/securelint/engine.py +213 -0
  29. securelint-1.0.0/src/securelint/explain.py +167 -0
  30. securelint-1.0.0/src/securelint/fingerprint.py +100 -0
  31. securelint-1.0.0/src/securelint/github_report.py +133 -0
  32. securelint-1.0.0/src/securelint/html_report.py +310 -0
  33. securelint-1.0.0/src/securelint/model.py +84 -0
  34. securelint-1.0.0/src/securelint/parsing.py +73 -0
  35. securelint-1.0.0/src/securelint/policy.py +57 -0
  36. securelint-1.0.0/src/securelint/prompt.py +132 -0
  37. securelint-1.0.0/src/securelint/providers.py +249 -0
  38. securelint-1.0.0/src/securelint/py.typed +0 -0
  39. securelint-1.0.0/src/securelint/report.py +172 -0
  40. securelint-1.0.0/src/securelint/rules/__init__.py +16 -0
  41. securelint-1.0.0/src/securelint/rules/_helpers.py +67 -0
  42. securelint-1.0.0/src/securelint/rules/_sources.py +121 -0
  43. securelint-1.0.0/src/securelint/rules/code_execution.py +66 -0
  44. securelint-1.0.0/src/securelint/rules/deserialization.py +112 -0
  45. securelint-1.0.0/src/securelint/rules/hardcoded_secrets.py +169 -0
  46. securelint-1.0.0/src/securelint/rules/shell_injection.py +112 -0
  47. securelint-1.0.0/src/securelint/rules/sql_injection.py +168 -0
  48. securelint-1.0.0/src/securelint/sarif_report.py +183 -0
  49. securelint-1.0.0/src/securelint/suppressions.py +169 -0
  50. securelint-1.0.0/src/securelint/trivy.py +308 -0
  51. securelint-1.0.0/tests/corpus/README.md +44 -0
  52. securelint-1.0.0/tests/corpus/code_execution.py +55 -0
  53. securelint-1.0.0/tests/corpus/deserialization.py +91 -0
  54. securelint-1.0.0/tests/corpus/hardcoded_secrets.py +57 -0
  55. securelint-1.0.0/tests/corpus/safe_app.py +49 -0
  56. securelint-1.0.0/tests/corpus/shell_injection.py +94 -0
  57. securelint-1.0.0/tests/corpus/sql_injection.py +95 -0
  58. securelint-1.0.0/tests/corpus/vulnerable_app.py +44 -0
  59. securelint-1.0.0/tests/test_action.py +516 -0
  60. securelint-1.0.0/tests/test_answers.py +201 -0
  61. securelint-1.0.0/tests/test_baseline.py +416 -0
  62. securelint-1.0.0/tests/test_benchmark.py +39 -0
  63. securelint-1.0.0/tests/test_cache.py +205 -0
  64. securelint-1.0.0/tests/test_changes.py +408 -0
  65. securelint-1.0.0/tests/test_cli.py +241 -0
  66. securelint-1.0.0/tests/test_compare_bandit.py +136 -0
  67. securelint-1.0.0/tests/test_config.py +141 -0
  68. securelint-1.0.0/tests/test_context.py +77 -0
  69. securelint-1.0.0/tests/test_corpus.py +84 -0
  70. securelint-1.0.0/tests/test_dataflow.py +262 -0
  71. securelint-1.0.0/tests/test_discovery.py +123 -0
  72. securelint-1.0.0/tests/test_engine.py +218 -0
  73. securelint-1.0.0/tests/test_explain.py +237 -0
  74. securelint-1.0.0/tests/test_fingerprint.py +182 -0
  75. securelint-1.0.0/tests/test_github_report.py +260 -0
  76. securelint-1.0.0/tests/test_html_report.py +202 -0
  77. securelint-1.0.0/tests/test_model.py +83 -0
  78. securelint-1.0.0/tests/test_parsing.py +123 -0
  79. securelint-1.0.0/tests/test_policy.py +328 -0
  80. securelint-1.0.0/tests/test_prompt.py +222 -0
  81. securelint-1.0.0/tests/test_providers.py +600 -0
  82. securelint-1.0.0/tests/test_real_projects.py +46 -0
  83. securelint-1.0.0/tests/test_release.py +162 -0
  84. securelint-1.0.0/tests/test_report.py +166 -0
  85. securelint-1.0.0/tests/test_rule_code_execution.py +83 -0
  86. securelint-1.0.0/tests/test_rule_deserialization.py +108 -0
  87. securelint-1.0.0/tests/test_rule_hardcoded_secrets.py +135 -0
  88. securelint-1.0.0/tests/test_rule_shell_injection.py +110 -0
  89. securelint-1.0.0/tests/test_rule_sources.py +192 -0
  90. securelint-1.0.0/tests/test_rule_sql_injection.py +99 -0
  91. securelint-1.0.0/tests/test_rules_follow_variables.py +328 -0
  92. securelint-1.0.0/tests/test_sarif_report.py +302 -0
  93. securelint-1.0.0/tests/test_settings_in_a_scan.py +176 -0
  94. securelint-1.0.0/tests/test_suppressions.py +347 -0
  95. securelint-1.0.0/tests/test_threat_model.py +45 -0
  96. securelint-1.0.0/tests/test_trivy.py +491 -0
@@ -0,0 +1,3 @@
1
+ # Store text files with LF line endings and check them out with LF on every
2
+ # OS, so Windows and Linux (CI) never disagree about line endings in diffs.
3
+ * text=auto eol=lf
@@ -0,0 +1,112 @@
1
+ name: CI
2
+
3
+ on:
4
+ push:
5
+ branches: [main]
6
+ pull_request:
7
+
8
+ # The workflow only reads the repository; it never needs to write to it.
9
+ permissions:
10
+ contents: read
11
+
12
+ # A newer push to the same branch or pull request replaces the older run.
13
+ concurrency:
14
+ group: ci-${{ github.ref }}
15
+ cancel-in-progress: true
16
+
17
+ jobs:
18
+ checks:
19
+ name: Checks (Python ${{ matrix.python-version }})
20
+ runs-on: ubuntu-latest
21
+ timeout-minutes: 10
22
+ strategy:
23
+ fail-fast: false
24
+ matrix:
25
+ # The oldest supported version and the newest released one.
26
+ python-version: ["3.11", "3.14"]
27
+ steps:
28
+ # Actions are pinned to a full commit SHA: a tag can be moved to
29
+ # different code, a commit cannot.
30
+ - name: Check out the repository
31
+ uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
32
+ with:
33
+ persist-credentials: false
34
+
35
+ - name: Set up Python ${{ matrix.python-version }}
36
+ uses: actions/setup-python@6d4fd22b20465ccc0f7354476588e8df7c9ca0e0 # v7.1.0
37
+ with:
38
+ python-version: ${{ matrix.python-version }}
39
+ cache: pip
40
+ cache-dependency-path: pyproject.toml
41
+
42
+ - name: Install the package and development tools
43
+ run: python -m pip install -e ".[dev]"
44
+
45
+ - name: Lint
46
+ run: python -m ruff check .
47
+
48
+ - name: Check formatting
49
+ run: python -m ruff format --check .
50
+
51
+ - name: Type check
52
+ run: python -m mypy
53
+
54
+ - name: Run tests
55
+ run: python -m pytest
56
+
57
+ securelint:
58
+ # SecureLint checks its own code with its own Action, the way another
59
+ # project would use it. "./" is the Action in this commit, so a change to
60
+ # action.yml is tested by the same pull request.
61
+ name: SecureLint
62
+ runs-on: ubuntu-latest
63
+ timeout-minutes: 10
64
+ steps:
65
+ - name: Check out the repository
66
+ uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
67
+ with:
68
+ fetch-depth: 0 # the whole history, to find the changed lines
69
+ persist-credentials: false
70
+
71
+ - name: Scan with SecureLint
72
+ uses: ./
73
+ with:
74
+ # Also download the pinned Trivy, verify it and read its report.
75
+ trivy: "true"
76
+
77
+ package:
78
+ # Builds the package the way it will be published, checks it as PyPI
79
+ # will, and installs it in an empty environment to use it from outside
80
+ # the repository, as someone who ran `pip install securelint` would.
81
+ name: Package
82
+ runs-on: ubuntu-latest
83
+ timeout-minutes: 10
84
+ steps:
85
+ - name: Check out the repository
86
+ uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
87
+ with:
88
+ persist-credentials: false
89
+
90
+ - name: Set up Python
91
+ uses: actions/setup-python@6d4fd22b20465ccc0f7354476588e8df7c9ca0e0 # v7.1.0
92
+ with:
93
+ python-version: "3.13"
94
+
95
+ - name: Build the package
96
+ run: |
97
+ python -m pip install "build==1.6.1" "twine==7.0.0"
98
+ python -m build
99
+ python -m twine check --strict dist/*
100
+
101
+ - name: Install it in an empty environment and use it
102
+ run: |
103
+ python -m venv "$RUNNER_TEMP/clean"
104
+ "$RUNNER_TEMP/clean/bin/python" -m pip install dist/*.whl
105
+ cd "$RUNNER_TEMP"
106
+ "$RUNNER_TEMP/clean/bin/securelint" --version
107
+ printf 'print("safe")\n' > ok.py
108
+ "$RUNNER_TEMP/clean/bin/securelint" scan ok.py
109
+ printf 'eval(input())\n' > risky.py
110
+ status=0
111
+ "$RUNNER_TEMP/clean/bin/securelint" scan risky.py || status=$?
112
+ test "$status" -eq 1
@@ -0,0 +1,116 @@
1
+ name: Release
2
+
3
+ # Publishes SecureLint with PyPI's Trusted Publishing: GitHub proves to PyPI
4
+ # which repository and workflow is asking, so no password or token is stored
5
+ # anywhere.
6
+ #
7
+ # - Started by hand (Actions > Release > Run workflow): a rehearsal that
8
+ # publishes to TestPyPI only.
9
+ # - A tag such as v1.0.0: publishes to TestPyPI, then to PyPI.
10
+ on:
11
+ workflow_dispatch:
12
+ push:
13
+ tags: ["v*"]
14
+
15
+ # Nothing here writes to the repository. The publishing jobs get the one
16
+ # extra permission they need, and nothing else does.
17
+ permissions:
18
+ contents: read
19
+
20
+ # Never two releases at once, and never one cancelled halfway.
21
+ concurrency:
22
+ group: release
23
+ cancel-in-progress: false
24
+
25
+ jobs:
26
+ build:
27
+ name: Build
28
+ runs-on: ubuntu-latest
29
+ timeout-minutes: 10
30
+ steps:
31
+ - name: Check out the repository
32
+ uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
33
+ with:
34
+ persist-credentials: false
35
+
36
+ - name: Set up Python
37
+ uses: actions/setup-python@6d4fd22b20465ccc0f7354476588e8df7c9ca0e0 # v7.1.0
38
+ with:
39
+ python-version: "3.13"
40
+
41
+ - name: Build and check the package
42
+ run: |
43
+ python -m pip install "build==1.6.1" "twine==7.0.0"
44
+ python -m build
45
+ python -m twine check --strict dist/*
46
+
47
+ - name: Check that the version matches the tag
48
+ env:
49
+ REF_TYPE: ${{ github.ref_type }}
50
+ TAG: ${{ github.ref_name }}
51
+ run: |
52
+ set -u
53
+ wheel=$(ls dist/*.whl)
54
+ version=$(basename "$wheel" | cut -d - -f 2)
55
+ echo "Package version: $version"
56
+ if [ "$REF_TYPE" = "tag" ] && [ "$TAG" != "v$version" ]; then
57
+ echo "::error::The tag $TAG does not match the package version $version"
58
+ exit 1
59
+ fi
60
+
61
+ - name: Keep the package for the publishing jobs
62
+ uses: actions/upload-artifact@cf430e030ddbb5b0abf93d22962f4752f3646cd9 # v7.0.2
63
+ with:
64
+ name: dist
65
+ path: dist/
66
+ if-no-files-found: error
67
+
68
+ testpypi:
69
+ name: Publish to TestPyPI
70
+ needs: build
71
+ runs-on: ubuntu-latest
72
+ timeout-minutes: 10
73
+ environment:
74
+ name: testpypi
75
+ url: https://test.pypi.org/p/securelint
76
+ permissions:
77
+ # Lets GitHub prove to TestPyPI who is publishing (Trusted Publishing).
78
+ id-token: write
79
+ steps:
80
+ - name: Get the package
81
+ uses: actions/download-artifact@9000827ccba6bdab643e8b6fd33ac0654aef8333 # v8.0.2
82
+ with:
83
+ name: dist
84
+ path: dist/
85
+
86
+ - name: Publish
87
+ # v1.14.3 fixes GHSA-5xjx-3w67-w9x9 (critical, October 2026).
88
+ uses: pypa/gh-action-pypi-publish@7245906fb61fa33a13554f9d5fb856fe61758086 # v1.14.3
89
+ with:
90
+ repository-url: https://test.pypi.org/legacy/
91
+ # A rehearsal can be run again for the same version.
92
+ skip-existing: true
93
+
94
+ pypi:
95
+ name: Publish to PyPI
96
+ # Only for a tag, and only after TestPyPI accepted the same files.
97
+ if: github.ref_type == 'tag'
98
+ needs: testpypi
99
+ runs-on: ubuntu-latest
100
+ timeout-minutes: 10
101
+ environment:
102
+ name: pypi
103
+ url: https://pypi.org/p/securelint
104
+ permissions:
105
+ # Lets GitHub prove to PyPI who is publishing (Trusted Publishing).
106
+ id-token: write
107
+ steps:
108
+ - name: Get the package
109
+ uses: actions/download-artifact@9000827ccba6bdab643e8b6fd33ac0654aef8333 # v8.0.2
110
+ with:
111
+ name: dist
112
+ path: dist/
113
+
114
+ - name: Publish
115
+ # v1.14.3 fixes GHSA-5xjx-3w67-w9x9 (critical, October 2026).
116
+ uses: pypa/gh-action-pypi-publish@7245906fb61fa33a13554f9d5fb856fe61758086 # v1.14.3
@@ -0,0 +1,16 @@
1
+ # Python bytecode
2
+ __pycache__/
3
+ *.py[cod]
4
+
5
+ # Virtual environment
6
+ .venv/
7
+
8
+ # Packaging output
9
+ build/
10
+ dist/
11
+ *.egg-info/
12
+
13
+ # Tool caches
14
+ .pytest_cache/
15
+ .ruff_cache/
16
+ .mypy_cache/
@@ -0,0 +1,248 @@
1
+ # Changelog
2
+
3
+ What changed in each version of SecureLint, newest first.
4
+
5
+ ## 1.0.0 (2026-11-16)
6
+
7
+ Measured, threat-modelled and installable: `pip install securelint`. The
8
+ README's Stability section lists what stays the same across 1.x (commands,
9
+ exit codes, JSON and SARIF formats, rule ids, fingerprints, settings and the
10
+ Action's inputs).
11
+
12
+ - `benchmarks/compare_bandit.py` runs SecureLint and Bandit 1.9.4 on the
13
+ annotated corpus and scores both the same way, by statement, for the five
14
+ kinds of problems both look for. The README shows the results and what
15
+ they leave out: the corpus favours SecureLint, and Bandit does better on
16
+ hard-coded secrets.
17
+ - `docs/real-projects.md` sorts by hand every finding of SecureLint and
18
+ Bandit on four real projects (Flask, Requests, HTTPie, Jinja):
19
+ SecureLint 6 findings with 1 false alarm, Bandit 20 with 15.
20
+ `benchmarks/real_projects.py` lists the findings again for any folder.
21
+ - `docs/threat-model.md` lists what SecureLint protects, what it does not
22
+ trust, the attacks it was built against with the test behind each
23
+ protection, and the remaining risks. A test fails if a named test
24
+ disappears. `SECURITY.md` says how to report a vulnerability privately.
25
+ - The Action installs Anthropic's package only when the project's settings
26
+ choose the `anthropic` provider, not for every `explain: "true"`.
27
+ - An OpenAI-compatible service that does not answer within 60 seconds is
28
+ asked once more, at once; a second slow answer ends the attempt.
29
+ - Ready for PyPI: the package now has classifiers, keywords and project
30
+ links, ships a `py.typed` marker, and the README's links are full
31
+ addresses so they also work on PyPI. A new CI job builds the package,
32
+ checks it with `twine check --strict`, and installs it in an empty
33
+ environment to use it from outside the repository.
34
+ - A Release workflow publishes with PyPI's Trusted Publishing (no password
35
+ or token stored anywhere) and digital attestations: by hand to TestPyPI
36
+ as a rehearsal, and on a `v*` tag to TestPyPI then PyPI, after checking
37
+ that the tag matches the package version. `docs/releasing.md` explains
38
+ the one-time setup and the steps of a release. A rehearsal published 0.6.0
39
+ to TestPyPI with attestations naming this repository and workflow.
40
+ - The README explains installing from PyPI and what is stable in 1.x.
41
+ - The corpus numbers are unchanged (54 reported, 9 missed, 4 reported by
42
+ mistake).
43
+
44
+ Known limits: the rules find five kinds of problems in Python code, not all
45
+ security problems; a pull request can silence its own findings, which is
46
+ visible but not blocked (see the threat model); there is no size limit on
47
+ scanned files.
48
+
49
+ ## 0.6.0 (2026-11-15)
50
+
51
+ AI explanations: a few plain words under each finding, from the AI the
52
+ project chooses, without ever changing what blocks.
53
+
54
+ - `scan --explain` adds a short explanation, written by an AI, under each
55
+ finding shown (at most 10), in the text and JSON reports. It is added
56
+ after the policy has decided, so it never changes what blocks or the exit
57
+ code, and an explanation that fails is only skipped.
58
+ - The AI is chosen in `[tool.securelint.ai]`: `anthropic` (Claude, through
59
+ Anthropic's Python package, an optional extra:
60
+ `pip install "securelint[anthropic]"`) or `openai-compatible` (OpenAI, or a
61
+ local model such as Ollama, with no package). The key is read from an
62
+ environment variable; a key written in the settings stops the scan. A
63
+ `base-url` must be `https://`, or `http://` on this machine, and redirects
64
+ are not followed. A busy OpenAI-compatible service (429 or 5xx) is asked
65
+ again up to two times, waiting longer each time.
66
+ - Explanations are kept in `.securelint-cache/` (never committed, no code in
67
+ it), so the same finding with the same lines around it is not sent to the
68
+ AI again. A damaged or unwritable cache is reported and changes nothing.
69
+ - The Action's `explain` input adds the explanations to the notes on a pull
70
+ request; the key comes from a repository secret given to the step. Only
71
+ the notes are explained, not the SARIF report, and pull requests from
72
+ forks, which get no secrets, are checked as usual without explanations.
73
+ - What is sent to explain a finding is the finding and the 3 lines before
74
+ and after it, never the whole file. Before it leaves, every string on the
75
+ line of a hard-coded secret, the value of any name that sounds secret, and
76
+ text shaped like a well-known key are replaced by `***`. Lines are cut at
77
+ 200 characters.
78
+ - The AI must answer with a JSON object holding an `explanation` and,
79
+ optionally, a `fix`; any other answer is dropped. Control characters are
80
+ removed and long answers are cut. Explanations are shown, labelled "AI
81
+ explanation (may be wrong)", in the text, JSON and HTML reports and in the
82
+ GitHub notes, never in the job summary.
83
+ - Tested for real with Gemini's free tier (`gemini-3.5-flash-lite`, through
84
+ the `openai-compatible` provider), on the laptop and in GitHub Actions on
85
+ [securelint-demo pull request #1](https://github.com/Fatennhrb/securelint-demo/pull/1).
86
+ The Claude provider is tested against Anthropic's real package and a fake
87
+ service, not yet with a real key.
88
+ - The corpus numbers are unchanged (54 reported, 9 missed, 4 reported by
89
+ mistake).
90
+
91
+ Known limits: the Action installs Anthropic's package whenever `explain` is
92
+ on, even for another provider; a request that times out is not retried; on
93
+ Gemini's free tier Google may use what is sent, so it suits made-up code only.
94
+
95
+ ## 0.5.0 (2026-11-15)
96
+
97
+ Dependencies and Dockerfiles: Trivy's findings go through the same gate as
98
+ the code's.
99
+
100
+ - `--trivy-report FILE` adds the vulnerable dependencies found by Trivy
101
+ (`trivy fs --format json`) to the scan, as findings of rule `SL201` on the
102
+ line that names the package. They go through the baseline, the policy and
103
+ every report like the other findings. SecureLint reads the report; it does
104
+ not run Trivy.
105
+ - The same report's failed configuration checks (Trivy's `misconfig`
106
+ scanner) become findings of rule `SL202`, on the Dockerfile line Trivy
107
+ points to.
108
+ - The Action's `trivy` input downloads Trivy 0.76.2, checks it against a
109
+ pinned checksum before running it, and adds its report to the scan.
110
+ SecureLint's own CI uses it, and
111
+ [securelint-demo pull request #2](https://github.com/Fatennhrb/securelint-demo/pull/2)
112
+ shows it on an old dependency and a Dockerfile.
113
+ - The summary counts the files a Trivy report has problems in among the
114
+ files checked.
115
+ - The corpus numbers are unchanged (54 reported, 9 missed, 4 reported by
116
+ mistake).
117
+
118
+ Known limits: SecureLint does not run Trivy itself outside the Action, and
119
+ does not scan built container images; a dependency is reported on the line
120
+ that names it, or on line 1 when no line does; "found in N files" counts the
121
+ files checked, not only those with problems.
122
+
123
+ ## 0.4.0 (2026-11-15)
124
+
125
+ GitHub integration: findings appear on the lines a pull request changed.
126
+
127
+ - `--format sarif` writes the result in SARIF 2.1.0, the format GitHub reads
128
+ for security tools, with paths from the top of the project and each
129
+ finding's fingerprint.
130
+ - `--changed-since BRANCH` reports only findings on lines the current
131
+ branch added or changed since it left BRANCH, asking Git for those
132
+ lines. Findings elsewhere are counted and listed, not reported.
133
+ - `--format github` prints findings as notes GitHub Actions shows on the
134
+ lines of a pull request, and adds a summary table to the job's page.
135
+ - The repository is a GitHub Action (`action.yml`): on a pull request it
136
+ reports the findings on the changed lines and fails when something
137
+ blocks. It needs no token and no write permission.
138
+ - SecureLint's own CI runs the Action on its code, and a demo repository,
139
+ [securelint-demo](https://github.com/Fatennhrb/securelint-demo), shows it
140
+ on a real pull request.
141
+ - The corpus numbers are unchanged (54 reported, 9 missed, 4 reported by
142
+ mistake).
143
+
144
+ Known limits: the Action does not upload SARIF itself; `--changed-since` needs
145
+ the target branch in the local history (`fetch-depth: 0`); a finding is shown
146
+ only when its own line changed, even if the change that caused it is on
147
+ another line.
148
+
149
+ ## 0.3.0 (2026-11-14)
150
+
151
+ Policy and configuration: what a team needs to use the scanner every day.
152
+
153
+ - Settings in `pyproject.toml`: `exclude` leaves files and directories out
154
+ of a scan, `ignore` turns rules off. A mistake in the settings stops the
155
+ scan instead of being skipped.
156
+ - One finding can be silenced with a comment on its line:
157
+ `# securelint: ignore[SL102] reason`. Silenced findings are counted and
158
+ listed in the reports, and a comment that silences nothing is reported as
159
+ `SL001`.
160
+ - Each finding has a `fingerprint` in the JSON report: an identity that
161
+ does not change when lines are added or removed elsewhere in the file.
162
+ - `securelint baseline` records the findings a project has today in
163
+ `.securelint-baseline.json`. Later scans report only new findings; the
164
+ known ones are counted, and listed in the JSON and HTML reports.
165
+ - A policy in `[tool.securelint.policy]` sets the severity and confidence
166
+ from which a finding blocks. Findings below it are reported as warnings
167
+ and do not fail the scan. Without a policy, every finding still blocks.
168
+ - The corpus numbers are unchanged (54 reported, 9 missed, 4 reported by
169
+ mistake), and the scan is not slower than in 0.2.0 when both are timed
170
+ side by side.
171
+
172
+ Known limits: `exclude` takes exact paths, not patterns; a silencing comment
173
+ must be on the line where the finding is reported; editing a line listed in
174
+ the baseline, even its comment, makes its finding new again.
175
+
176
+ ## 0.2.0 (2026-11-13)
177
+
178
+ Data-flow analysis: the rules follow a value from where it is created to
179
+ where it is used, inside one file.
180
+
181
+ - Rules look through a variable to the value it was given, inside one
182
+ function. A query built on one line and run on the next is now reported,
183
+ and a fixed value kept in a variable is no longer reported. This is done
184
+ only when the variable is assigned exactly once.
185
+ - A function kept in a variable is followed: after `run = os.system`,
186
+ `run(command)` is checked as `os.system(command)`.
187
+ - A value set once at the top of a file is known inside the functions of
188
+ that file, and fixed pieces put together (`"ls " + "-l"`, an f-string of
189
+ fixed values, `", ".join("?" for _ in ids)`) count as fixed. This removes
190
+ false alarms.
191
+ - A finding whose value comes from a web request, `input()`, `sys.argv` or
192
+ `sys.stdin` is reported with high confidence, and its message says where
193
+ the value comes from and on which line.
194
+ - Corpus: 54 dangerous lines reported (was 50), 9 missed (was 13), 4 safe
195
+ lines reported by mistake (was 8).
196
+ - Speed: 314 files per second on Python's standard library (was 299).
197
+
198
+ Known limits: a value passed from one function to another is not followed,
199
+ and a variable assigned more than once is not followed.
200
+
201
+ ## 0.1.0 (2026-11-12)
202
+
203
+ The first release: a working scanner for Python source code.
204
+
205
+ ### Scanning
206
+
207
+ - `securelint scan [path]` scans one Python file, or every Python file in a
208
+ directory. Hidden directories, `__pycache__`, `node_modules`, virtual
209
+ environments and symbolic links are skipped.
210
+ - Files are only parsed, never imported or run. A file that cannot be read or
211
+ parsed is reported and the scan continues with the others.
212
+ - Exit codes for scripts and CI jobs: `0` no problems, `1` problems found,
213
+ `2` the scan could not be completed.
214
+
215
+ ### Rules
216
+
217
+ - **SL101**: `eval()` or `exec()` called with a value built at run time.
218
+ - **SL102**: a shell command built at run time (`os.system()`,
219
+ `subprocess` with `shell=True`, and similar).
220
+ - **SL103**: data loaded with a format that can run code (`pickle`, `dill`,
221
+ `marshal`, `yaml.load()` without a safe loader).
222
+ - **SL104**: a password or key written directly in the code.
223
+ - **SL105**: an SQL query with values put into its text.
224
+ - Rules follow imports, so `import subprocess as sp` and
225
+ `from os import system as sh` are understood.
226
+
227
+ ### Reports
228
+
229
+ - Text report for the terminal (the default).
230
+ - `--format json`: the full result as a JSON document.
231
+ - `--format html`: one self-contained page to open in a browser, with no
232
+ script and no network access.
233
+
234
+ ### Quality
235
+
236
+ - An annotated corpus of example files measures the rules: 50 dangerous
237
+ lines reported, 13 missed, 8 safe lines reported by mistake.
238
+ - A benchmark script measures scan speed: 299 files per second on Python's
239
+ standard library, on the author's laptop.
240
+ - Every change is checked on Python 3.11 and 3.14 with Ruff, mypy (strict)
241
+ and pytest.
242
+
243
+ ### Known limits
244
+
245
+ - A value that passes through a variable before reaching a dangerous call is
246
+ not followed (`query = f"..."` then `cursor.execute(query)`).
247
+ - `from module import *` is not followed.
248
+ - There is no configuration file and no way to silence a finding yet.
@@ -0,0 +1,21 @@
1
+ MIT License
2
+
3
+ Copyright (c) 2026 Faten Harbaoui
4
+
5
+ Permission is hereby granted, free of charge, to any person obtaining a copy
6
+ of this software and associated documentation files (the "Software"), to deal
7
+ in the Software without restriction, including without limitation the rights
8
+ to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
9
+ copies of the Software, and to permit persons to whom the Software is
10
+ furnished to do so, subject to the following conditions:
11
+
12
+ The above copyright notice and this permission notice shall be included in all
13
+ copies or substantial portions of the Software.
14
+
15
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16
+ IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17
+ FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
18
+ AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19
+ LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
20
+ OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
21
+ SOFTWARE.