AppSec Rules Pack

Reusable AppSec policy-as-code rules for secure application review, CI quality gates, and manual evidence collection.

This initial pack is intentionally generic. It does not contain product names, tenant identifiers, customer data, secrets, internal endpoints, or environment-specific configuration.

Why / When not to use

Use this rule contract and engine-agnostic validator to keep AppSec review guidance, expected evidence, remediation, and exceptions consistent across teams and CI gates. It validates the structure and consistency of rule packs.

It does not execute rules or scan application code. Use a separate scanner when you need vulnerability findings. Framework mappings support review; they do not establish compliance. The derived Semgrep scaffold has placeholder patterns, and the SARIF catalog has no findings. See the scope and ADR-0001.

The demo records real CLI output. Recreate it from a source checkout with python docs/assets/record-cli-demo.py after installing the project.

What Is Included

• A short technical specification in TECHNICAL_SPEC.md and a direction summary in ROADMAP.md.
• A JSON Schema rule contract in src/appsec_rules_pack/schemas/appsec-rule.schema.json.
• A baseline YAML rules pack of 19 generic rules in rules/appsec-baseline.yaml, covering authentication, authorization, input validation, injection (including output-encoding/XSS), SSRF, secrets, file handling, logging, dependency risk, configuration, session hardening, CSRF, webhook/message integrity, excessive data exposure, mass assignment, open redirect, and rate limiting. Every rule ships an explicit compliant and violating code example.
• A Python 3.12+ validator with a Typer CLI supporting --version, --fail-on-warnings, --require-examples, and --format json output for CI, plus derivation-only export index, export semgrep, export sarif, and report coverage subcommands.
• Derived, drift-tested artifacts under exports/: a machine-readable rule index (appsec-baseline.index.json), a clearly labeled NON-runnable Semgrep scaffold, and a SARIF 2.1.0 rule catalog (empty results). Derivation only — the validator stays engine-agnostic and never executes rules (ADR-0001).
• Unit tests and pass/fail/warn fixtures for valid packs, invalid schema shape, enum/type/additionalProperties failures, duplicate rule IDs, cross-file duplicate IDs, exception-window warnings, exception-policy contradictions, malformed framework mapping IDs, and sensitive-value detection.
• A 95% coverage gate plus a hardened CI/CD surface: a build/lint/test workflow (with separate jobs running the suite on Ubuntu and Windows, and on Python 3.13), a security pipeline (Semgrep, CodeQL, Bandit, Trivy, KICS, pip-audit, Gitleaks, Dependency Review, actionlint), and OpenSSF Scorecard analysis. The build/lint/test, cross-platform, and security jobs are required status checks on master.
• Architecture decision records in docs/adr/, including the two that define this project's boundary: the validator stays engine-agnostic (ADR-0001) and the CI gate consumes its JSON rather than embedding enforcement (ADR-0004).
• Contribution guidance for safe rule additions, a code of conduct, and issue/PR templates.
• A CI integration template in examples/.

Project Layout

.
|-- .github/
|   |-- ISSUE_TEMPLATE/     # bug reports, rule proposals, and template configuration
|   |-- workflows/         # CI, security, policy gate, Scorecard, and publishing
|   |-- CODEOWNERS
|   |-- PULL_REQUEST_TEMPLATE.md
|   `-- dependabot.yml
|-- docs/
|   |-- adr/               # architecture decision records
|   `-- assets/            # recorded CLI demo and its reproduction script
|-- examples/
|   `-- README.md          # GitHub Actions integration example
|-- exports/
|   |-- appsec-baseline.index.json
|   |-- sarif/appsec-baseline.sarif.json
|   `-- semgrep/appsec-baseline.semgrep.yaml
|-- rules/appsec-baseline.yaml
|-- src/appsec_rules_pack/
|   |-- __init__.py
|   |-- __main__.py
|   |-- cli.py
|   |-- exporter.py
|   |-- loader.py
|   |-- reporter.py
|   |-- sarif_export.py
|   |-- semgrep_scaffold.py
|   |-- validator.py
|   `-- schemas/appsec-rule.schema.json
|-- tests/
|   |-- fixtures/          # valid, invalid, warning, and multi-file packs
|   |-- helpers.py
|   `-- test_*.py
|-- CHANGELOG.md
|-- CODE_OF_CONDUCT.md
|-- CONTRIBUTING.md
|-- LICENSE
|-- README.md
|-- ROADMAP.md
|-- SECURITY.md
|-- STATUS.md
|-- TECHNICAL_SPEC.md
`-- pyproject.toml

Installation

Install the published package from PyPI:

pip install appsec-rules-pack

This installs the appsec-rules console script. Pin to a reviewed version (for example appsec-rules-pack==0.4.0) when using it in a CI quality gate.

What the distribution contains: the validator, the CLI, and the JSON Schema. It does not ship a rules pack — the CLI validates whatever path you point it at.

To get the baseline pack of 19 rules, either take the version-pinned copy attached to each GitHub Release:

curl -LO https://github.com/lucashgrifoni/AppSec-Rules-Pack/releases/download/v0.4.0/appsec-baseline.yaml

or copy rules/appsec-baseline.yaml from this repository. You can also write your own pack from the example below.

Verify a release

After downloading a release asset, verify its SLSA build-provenance attestation with a current GitHub CLI:

gh attestation verify appsec-baseline.yaml --repo lucashgrifoni/AppSec-Rules-Pack

For a tighter check, also require the publishing workflow and the release tag you chose:

gh attestation verify appsec-baseline.yaml \
  --repo lucashgrifoni/AppSec-Rules-Pack \
  --signer-workflow lucashgrifoni/AppSec-Rules-Pack/.github/workflows/publish-pypi.yml \
  --source-ref refs/tags/v0.4.0

From v0.4.0 each Release also carries the signed provenance bundle as appsec-rules-pack-<tag>.intoto.jsonl, so the check also works without querying the attestation API:

gh attestation verify appsec-baseline.yaml \
  --repo lucashgrifoni/AppSec-Rules-Pack \
  --bundle appsec-rules-pack-v0.4.0.intoto.jsonl

Use the same command with a downloaded wheel, source distribution, or sbom.cdx.json. The publishing workflow attests those files and the baseline pack; ADR-0003 explains the release design. A successful check verifies the artifact against its signed provenance and the selected identity constraints. It does not certify that the rule guidance is complete or that an application is secure. Treat a failed or unavailable verification as unverified provenance.

From source (development)

python -m venv .venv
.\.venv\Scripts\python -m pip install -e ".[dev]"

On macOS/Linux:

python -m venv .venv
.venv/bin/python -m pip install -e ".[dev]"

If the dependencies already exist in the active Python environment, the validator can also be run directly with PYTHONPATH=src.

Your First Rules Pack

A pack is a pack block plus one or more rules. This is a complete, valid minimal pack — copy it, run appsec-rules validate on it, then grow it. Every field shown is required; the full contract is in appsec-rule.schema.json, and the 19 baseline rules in rules/appsec-baseline.yaml are worked examples.

pack:
  id: my-pack
  name: My Rules Pack
  version: 0.1.0
  mode: advisory
  owner: appsec
  description: A minimal rules pack to start from.

rules:
  - id: APPSEC-EXAMPLE-001
    title: Require controlled error handling
    description: Verify that invalid user input returns controlled errors without stack traces.
    severity: medium
    category: configuration
    status: enabled
    enforcement: advisory
    targets:
      - api
    mappings:
      owasp_asvs:
        - V14.4
      owasp_api_top_10_2023:
        - API8:2023
      cwe:
        - CWE-209
      nist_ssdf:
        - PW.7
    evidence:
      required:
        - Error handling path returns a documented response shape.
      signals:
        - Negative tests assert controlled error messages for invalid input.
    match:
      type: review
      includes:
        - API handlers that process untrusted input.
      excludes:
        - Local developer-only scripts that are not shipped.
    remediation:
      guidance: Replace raw exception output with a controlled error contract.
      validation:
        - Run negative tests for invalid input and malformed requests.
    exceptions:
      allowed: true
      max_days: 30
      required_fields:
        - owner
        - justification
        - expires_at

Note the shapes that are easy to guess wrong: evidence and match are objects (not lists), remediation needs guidance, and required_fields accepts only owner, justification, expires_at, compensating_control, and validation_plan.

Usage

The examples below use rules/appsec-baseline.yaml, which exists in this repository. If you installed from PyPI, point the CLI at your own rules file or directory instead.

Validate the baseline rules pack:

python -m appsec_rules_pack validate rules/appsec-baseline.yaml

Validate every .yaml or .yml rules pack under a directory:

python -m appsec_rules_pack validate rules

Or, after installation, use the console script:

appsec-rules validate rules/appsec-baseline.yaml

Fail on warnings as well as errors:

appsec-rules validate rules/appsec-baseline.yaml --fail-on-warnings

Warn when an enabled rule ships no compliant/violating examples (opt-in):

appsec-rules validate rules --require-examples

Emit machine-readable JSON for CI pipelines:

appsec-rules validate rules --format json

Show the installed version:

appsec-rules --version

Derive a machine-readable rule index (JSON) for downstream tooling. This only reads and derives pack metadata; it does not execute rules or emit findings:

appsec-rules export index rules/appsec-baseline.yaml
appsec-rules export index rules/appsec-baseline.yaml --output exports/appsec-baseline.index.json

Derive the non-executable Semgrep scaffold. It contains rule metadata and TODO-REPLACE-WITH-DETECTION-PATTERN placeholders, not detection logic:

appsec-rules export semgrep rules/appsec-baseline.yaml --output exports/semgrep/appsec-baseline.semgrep.yaml

Derive a SARIF 2.1.0 rule catalog. Its tool.driver.rules describes the rules, and its results array is empty because no scanner ran:

appsec-rules export sarif rules/appsec-baseline.yaml --output exports/sarif/appsec-baseline.sarif.json

Report how many rules have a mapping for each framework, plus missing rule IDs and counts by AppSec category. This is mapping coverage, not security-test or compliance coverage:

appsec-rules report coverage rules/appsec-baseline.yaml
appsec-rules report coverage rules/appsec-baseline.yaml --format json --output coverage.json

Export and coverage commands derive metadata without running the validator first. Run validate rules --require-examples --fail-on-warnings separately as the quality gate.

The JSON validation report (validate --format json) contains a summary object (files, rules, errors, warnings, ok) and a files array with per-file issues (level, path, message). Validation exits non-zero on errors or on warnings when --fail-on-warnings is set, matching the text output.

Use It In Your CI

examples/README.md contains a GitHub Actions template for installing the pack from PyPI, validating a rules directory, and failing the build on errors (and optionally warnings). Pin the install to a reviewed release version before enabling it as a quality gate.

Rule Pack Model

Rules are advisory by default. Each rule defines:

• a stable ID and severity;
• the target surface and AppSec category;
• framework mappings such as OWASP ASVS 5.0, OWASP API Security Top 10, OWASP Top 10:2025 (optional), CWE, and NIST SSDF;
• expected evidence and review signals;
• match guidance for reviewers or automation;
• remediation and validation guidance;
• exception metadata requirements;
• a compliant and a violating code example with a short explanation.

The initial version is optimized for reviewability and deterministic validation, not for deep scanner-specific matching.

Validation

python -m pytest
PYTHONPATH=src python -m appsec_rules_pack validate rules

The validator checks JSON Schema compliance, duplicate rule IDs within a file and across a validated directory, exception-window limits, exception-policy contradictions (for example, a disallowed exception that still declares a window), malformed framework mapping identifiers (CWE, OWASP API Top 10 2023, OWASP ASVS, NIST SSDF), and basic sensitive-value patterns. Directory validation reports each issue with the relative file path, schema path, severity, and a concise remediation-oriented message.

Contributing

See CONTRIBUTING.md for rule authoring principles, the severity model, exception requirements, and the required checks. All participation is governed by CODE_OF_CONDUCT.md. To report a security issue, follow SECURITY.md.

License

Licensed under the Apache License 2.0. See LICENSE.

Optional executable Semgrep references

The hand-maintained Semgrep reference layer contains two tested Python/Flask detections:

• APPSEC-INJECT-001: Flask query/form values reaching locally established sqlite3 SQL arguments; parameterized values are negative fixtures
• APPSEC-SSRF-001: Flask query/form values reaching module-level Requests URL arguments; fixed destination URLs are negative fixtures

The other 17 baseline rule IDs have no executable detection. Even these two detections cover only their documented source/sink combinations, with known false positives and false negatives. They do not establish complete coverage of SQL injection or SSRF. See the layer README for the full coverage inventory and limitations.

Install Semgrep separately and run the fixture suite from a source checkout:

python -m pip install "semgrep==1.179.0"
semgrep --test --strict --metrics=off --config exports/semgrep-rules/rules exports/semgrep-rules/tests

To scan your own Python source, point that separately installed engine at the reference rules and your source directory. The appsec-rules CLI does not invoke Semgrep, and export semgrep continues to emit the non-executable metadata scaffold. ADR-0005 records this boundary.