Dependency Risk Audit

Dependency Risk Audit

Run a repeatable dependency-risk audit for Python projects and return a prioritized remediation plan.

Workflow

Step 1: Identify dependency source of truth

1. Detect package manager and files in this order:
   • poetry.lock + pyproject.toml
   • uv.lock + pyproject.toml
   • Pipfile.lock + Pipfile
   • requirements*.txt and optional constraints*.txt
2. Prefer lockfiles for resolved versions.
3. Record Python runtime constraint from:
   • pyproject.toml (requires-python)
   • .python-version
   • CI config (if present)

Step 2: Build dependency inventory

Create an inventory with:

• package name
• installed/resolved version
• dependency type (direct or transitive)
• pin style (exact, range, unbounded)
• marker constraints (Python version, platform markers)

If direct/transitive split is unavailable from project files, state that limitation explicitly.

Step 3: Run security advisory checks

Prefer pip-audit. If unavailable, fall back to lockfile/static analysis and mark advisories as partially evaluated.

Common commands:

# requirements.txt projects
pip-audit -r requirements.txt -f json

# active environment
pip-audit -f json

For non-requirements workflows, export to requirements format first when possible, then audit that export.

For each finding, capture:

• advisory ID (for example PYSEC-*, CVE-*, GHSA-*)
• package + affected version
• fixed version range
• severity (if available)
• exploit or impact notes (if available)

Step 4: Detect stale pins

Classify each direct dependency:

• current: no newer release in same major
• minor/patch stale: behind within same major
• major stale: newer major available
• unknown: latest data unavailable

Flag stale pins as higher risk when:

• package is internet-facing, auth-related, crypto-related, or framework/core runtime
• current version is multiple majors behind
• package has a recent advisory history

Step 5: Evaluate upgrade-path safety

For each dependency requiring change, assess upgrade risk:

• low: patch/minor upgrade, no known breaking changes
• medium: minor upgrade with behavior/config changes
• high: major upgrade, Python-version jump, or resolver conflicts likely

Check these risk signals:

• major-version jump required to reach fixed secure version
• declared Python requirement of target version conflicts with project runtime
• conflicting upper bounds across dependencies
• framework-coupled libraries that typically require code migrations

When possible, propose a stepwise path:

1. patch/minor upgrades first
2. isolated major upgrades next
3. framework/runtime upgrades last

Step 6: Produce the report

Return a markdown report using this structure:

## Dependency Risk Audit

Audit root: `<path>`
Dependency source: `<lockfile or manifest>`
Python runtime target: `<version/constraint>`

### Executive Summary

- Overall risk: [LOW|MEDIUM|HIGH|CRITICAL]
- Known advisories: X (Critical Y, High Z, ...)
- Stale direct dependencies: X (Major-stale Y)
- Unsafe upgrade paths: X

### Security Advisories

| Package | Version | Advisory | Severity | Fixed In | Notes |
| ------- | ------- | -------- | -------- | -------- | ----- |

### Stale Pins

| Package | Current | Latest | Drift Type | Risk Notes |
| ------- | ------- | ------ | ---------- | ---------- |

### Upgrade Path Risks

| Package | Current -> Target | Risk | Why Risky | Recommended Path |
| ------- | ----------------- | ---- | --------- | ---------------- |

### Prioritized Remediation Plan

1. ...
2. ...
3. ...

### Not Evaluated

- Item + reason

Scoring Guidance (optional)

If the user asks for a numeric score, start at 100 and deduct:

• Critical advisory: -15
• High advisory: -10
• Medium advisory: -6
• Low advisory: -3
• Major-stale direct dependency: -4
• High-risk upgrade path: -5

Floor at 0. Cap repeated deductions for the same package/advisory pair.

Remediation Loop

If the user asks for fixes:

1. Address critical/high advisories first, prioritizing low-risk upgrades.
2. Re-run audit steps 2-6.
3. Report risk delta, unresolved blockers, and required code changes/tests.