Skip to content
Back to skills

Ops

ASecurity

Define a dependency lifecycle policy — patch/minor/major upgrade cadence, SBOM refresh, CVE triage, EOL tracking, automated PRs, lockfile hygiene, and supply-chain pinning.

  • 23 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added October 14, 2026
ai-agentsgorailssecurity

Security analysis

A100/100

Pro scans all 21 files and shows the line behind each finding

Scanned October 14, 2026

npx -y skills add andreibesleaga/GABBE --skill ops --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Ops?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Ops
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/andreibesleaga-ops-7f09e089/badge)](https://www.skillsdirectory.com/skills/andreibesleaga-ops-7f09e089)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: dependency-lifecycle
description: Define a dependency lifecycle policy — patch/minor/major upgrade cadence, SBOM refresh, CVE triage, EOL tracking, automated PRs, lockfile hygiene, and supply-chain pinning.
triggers:
  - define our dependency upgrade cadence
  - set up a dependency lifecycle policy
  - triage these cve advisories
  - refresh the sbom for the next release
  - track end-of-life dependencies
  - configure automated dependency pull requests
  - enforce lockfile and pinning hygiene
tags: [ops, dependencies, supply-chain, maintenance]
core: false
context_cost: medium
---
# Dependency Lifecycle Skill

## Goal
Produce a dependency lifecycle policy that keeps a project's third-party dependencies current, supported, and free of known vulnerabilities without churn-driven breakage. It sets the cadence for absorbing patch, minor, and major updates, keeps the SBOM and CVE posture fresh, tracks end-of-life (EOL) before it bites, and enforces lockfile and pinning hygiene. This complements `dependency-security.skill` (which owns SBOM generation and SCA scanning mechanics) by defining the ongoing *operational rhythm* around it.

## Steps
1. **Inventory and classify.** Enumerate direct and transitive dependencies and classify each by blast radius (runtime-critical, build-only, dev-only) and update risk. Distinguish **patch** (bug/security, low risk — adopt fast), **minor** (backward-compatible features, moderate risk — batched), and **major** (breaking, high risk — planned with migration work).
2. **Set the cadence.** Define the rhythm per class: patch/security as soon as CI passes; minor on a regular batched window (e.g., weekly/biweekly); major as scheduled, owned migration tasks. State who reviews and merges each class.
3. **Refresh the SBOM.** Tie SBOM regeneration to the release pipeline so every shipped artifact has a current bill of materials. Diff the SBOM between releases to surface newly introduced or removed components. Defer generation mechanics to `dependency-security.skill`.
4. **Triage CVEs.** When an advisory lands, assess: is the vulnerable package actually used, is the vulnerable code path reachable, and what is the severity/exploitability in context. Prioritize remediation (upgrade, patch, or compensating control) by real risk, not raw CVSS alone. Record decisions in the audit trail.
5. **Track EOL.** Maintain a list of dependencies (and runtimes/base images) approaching end-of-support, with the EOL date and the planned migration target. EOL software stops receiving security fixes — schedule the move before the date, not after.
6. **Automate dependency PRs.** Configure automated update PRs (grouped by class to reduce noise), with CI gates that must pass before merge. Auto-merge is permissible only for patch/dev updates with green CI and human review reserved for minor/major.
7. **Enforce lockfile hygiene.** Require committed, deterministic lockfiles; every dependency change updates the lockfile in the same PR. Reject PRs that change manifests without the corresponding lockfile update. Verify resolved versions match the lockfile in CI.
8. **Pin the supply chain.** Pin dependencies (and CI actions/base images) to exact versions or integrity hashes/digests rather than floating ranges or mutable tags, so builds are reproducible and a hijacked upstream tag cannot silently change resolved code.
9. **Assign ownership and write the policy.** Name owners for each class and for EOL tracking, then assemble the Output Format below.

## Constraints
- Patch/security updates are adopted promptly; majors are always planned migration work, never auto-merged.
- The SBOM is regenerated for every release artifact, not occasionally.
- CVE triage prioritizes by reachability and exploitability in context, not CVSS score alone, and records the decision.
- Lockfiles are committed, deterministic, and updated atomically with manifest changes.
- Dependencies, CI actions, and base images are pinned to exact versions or digests; mutable tags are not acceptable for reproducible builds.
- This skill defines policy and proposes PRs; it does not auto-merge minor/major changes or override CI gates without human approval.

## Output Format
A dependency lifecycle policy in Markdown:
- **Inventory & classification** — dependency classes (runtime/build/dev) and update-risk tiers.
- **Cadence** — table of patch/minor/major rhythm, reviewer, merge rule.
- **SBOM** — regeneration trigger and diff process (cross-link `dependency-security.skill`).
- **CVE triage** — assessment criteria and prioritization rule.
- **EOL tracking** — table of dependency, EOL date, migration target, owner.
- **Automation** — PR grouping, CI gates, auto-merge boundaries.
- **Lockfile & pinning** — hygiene rules and pinning policy (versions/digests).
- **Owners** — named owners per class and for EOL.

## Security & Guardrails

### 1. Skill Security
- **Risk**: Adopting a malicious or typosquatted package during an "upgrade." Mitigation: verify package provenance/integrity hashes, restrict installs to approved registries, and review new transitive additions in dependency PRs.
- **Risk**: A CVE dismissed as "not exploitable" without evidence, leaving a real hole. Mitigation: require a documented reachability/exploitability justification recorded in the audit trail for any advisory marked as accepted-risk.

### 2. System Integration Security
- **Risk**: Floating version ranges or mutable CI action tags letting a hijacked upstream inject code into the build. Mitigation: pin to exact versions or commit/digest hashes and verify integrity in CI; treat un-pinning as a security-reviewed change.
- **Risk**: Auto-merge bots merging a poisoned or breaking update with weak CI. Mitigation: gate all auto-merge on the full CI suite plus SCA scan; restrict auto-merge to patch/dev classes with required status checks under branch protection.

### 3. LLM & Agent Guardrails
- **Risk**: The agent bumping a major version and "fixing" the breakage by deleting tests or loosening assertions. Mitigation: major upgrades are human-reviewed migration tasks; the agent may not weaken test coverage to make an upgrade pass.
- **Risk**: The agent suppressing or down-ranking a security advisory to keep a PR green. Mitigation: CVE severity may not be edited by the agent; unresolved advisories above the policy threshold block merge and escalate to a human.

Files in this skill

  • caching-strategy.skill.md3.9 KB
  • capacity-planning.skill.md3.6 KB
  • cloud-deploy.skill.md3.2 KB
  • configuration-management.skill.md7.6 KB
  • cost-optimization.skill.md2.3 KB
  • decommission-sunset.skill.md6.6 KB
  • dependency-lifecycle.skill.md6.3 KB
  • deployment.skill.md7.8 KB
  • dev-environments.skill.md1.9 KB
  • dev-workflow.skill.md1.8 KB
  • docker-dev.skill.md3.1 KB
  • enterprise-integration.skill.md3.1 KB
  • feature-flag-management.skill.md5.7 KB
  • incident-response.skill.md2.8 KB
  • infra-devops.skill.md3.2 KB
  • k8s-dev.skill.md3.2 KB
  • memory-optimization.skill.md2.9 KB
  • observability-stack-setup.skill.md6.6 KB
  • performance-audit.skill.md7.1 KB
  • performance-optimization.skill.md3.1 KB

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.