kirvin

secure-sdlc

Security reference library for this stack. Covers threat modeling (STRIDE), security requirements (EARS), pre-merge security review checklist, and incident response playbooks. Load when designing integrations, reviewing security-sensitive PRs, or responding to incidents.

kirvin 0 Updated 5mo ago

Resources

1
GitHub

Install

npx skillscat add kirvin/agent-dev-harness/secure-sdlc

Install via the SkillsCat registry.

SKILL.md

Secure SDLC

A reference library for security work in this stack. Load the relevant reference for
the current task rather than reading everything upfront.

When to load each reference

Task Reference
Threat modeling a new feature or integration references/threat-model.md
Writing EARS security invariants references/security-requirements.md
Pre-merge security review of a PR references/pre-merge-checklist.md
Responding to a credential leak or incident references/incident-response.md

Quick orientation

Threat model → identify what could go wrong (STRIDE) → feeds into EARS + ADR layers

Security requirements → EARS invariants that encode non-negotiable security constraints;
one file per domain (auth.md, data.md, pipeline.md)

Pre-merge checklist → runnable commands + human-review items; must pass before merging
security-sensitive changes

Incident response → step-by-step playbooks for AWS credential compromise, Figma token
leak, git history exposure; each playbook is self-contained

Stack context

This skill is calibrated for the agent-dev-harness stack:

  • AWS Bedrock as the Claude Code provider (SSO-based auth via AWS profiles)
  • GitHub Actions CI with SHA-pinned actions
  • Figma API integration (personal access tokens)
  • Beads issue tracker for finding remediation
  • Node.js / TypeScript projects

Relationship to other security skills

  • kf:security-sweep — retrospective analysis after code is written; this skill is for during design and review
  • ce:managing-pipelines — CI/CD hardening; complements this skill for pipeline-specific changes

Load the reference that matches your current task. Do not load all four unless the situation requires it.

Categories