oss-license-review

name	oss-license-review
description	Analyze open-source license terms and compliance obligations for a repository. Use when asked to identify project licenses, explain what is required when reusing code, assess copyleft/commercial compatibility, or produce a practical risk checklist before shipping.

OSS License Review

Perform a practical open-source license review from repository evidence and translate findings into deploy-time obligations.

Workflow

1. Collect evidence first, then explain obligations.
2. Distinguish facts from inference:
   • Facts: detected files, SPDX identifiers, declared licenses.
   • Inference: likely obligations and compatibility risk.
3. Report uncertainty explicitly when license evidence is missing or contradictory.
4. State that this is compliance guidance, not legal advice.

Step 1: Gather License Evidence

Run the inventory script on the target repo:

python3 scripts/license_inventory.py /path/to/repo --use-case saas --modified

If needed, output machine-readable JSON:

python3 scripts/license_inventory.py /path/to/repo --use-case binary --json

Usage scenario flags:

• --use-case unknown|internal|saas|binary|source
• --modified when you changed upstream code

Always inspect at least:

• Root LICENSE*, COPYING*, NOTICE*
• package.json, pyproject.toml, Cargo.toml, pom.xml, composer.json (if present)
• package-lock.json for dependency license hints (if present)
• sbom.json, cyclonedx.json, spdx.json, *.spdx.json (if present)

The script evaluates SPDX expressions (AND / OR / WITH / parentheses), models a broad SPDX exception map (40+ entries) in risk scoring, and performs exception/base-license compatibility checks. It reports branch risk range (min..max) and warnings when exception compatibility is uncertain. For very complex expressions, evaluation applies depth/branch limits and reports truncation warnings. For large SBOM files, the parser uses streaming mode when optional ijson is available, and falls back to full JSON parsing on stream failure.

Step 2: Classify License Type and Risk

Use references/license-obligations.md to map detected licenses to obligations.

Risk baseline:

• low: Permissive licenses (MIT/BSD/Apache-2.0) with clear attribution path.
• medium: Weak copyleft or file-level copyleft (MPL/LGPL) or unclear evidence.
• high: Strong/network copyleft (GPL/AGPL), license conflict, or missing license.

Compatibility checks to always perform:

• Planned usage mode: internal-only, SaaS, binary distribution, source redistribution.
• Whether code is modified and redistributed.
• Whether static/dynamic linking or derivative work risk exists.
• Patent and notice requirements (especially Apache-2.0 and NOTICE files).

Step 3: Produce Actionable Output

Return results in this structure:

License Findings
- Primary project license: <id or unknown>
- Additional detected licenses: <list>
- Evidence: <files and key declarations>

Usage Impact
- Your scenario: <internal / SaaS / distributed binary / source release>
- Required actions: <attribution, notice, source disclosure, license copy, changes disclosure>
- Restrictions and conflict notes: <if any>

Risk Assessment
- Level: <low/medium/high>
- Why: <short rationale>
- Unknowns: <missing files, ambiguous statements, dual-license uncertainty>

Respect and Give Back
- Suggested actions: <upstream fixes, docs contribution, clear changelog, sponsorship>

Recommended Next Actions
1. <action>
2. <action>
3. <action>

Rules for Answers

• Prefer precise file references over generic claims.
• Do not claim full transitive dependency compliance unless fully verified.
• If data is incomplete, provide a conservative checklist and list missing evidence.
• Flag AGPL/GPL usage for distribution or hosted-service scenarios.
• Mention trademark and patent caveats when relevant.
• Include both mandatory compliance actions and optional give-back actions.
• For SPDX composite expressions (e.g., MIT OR Apache-2.0), state the branch-selection assumption explicitly.
• For CycloneDX 1.5+ SBOMs, inspect license fields from metadata, components, services, and evidence.