skills.rest
Jin9
Jin9COMMUNITY
Explore all skills by this author →
GitHub

red-teaming-implementation-plan

Adversarially review implementation plans and emit severity-ranked PROCEED, REVISE, or BLOCK verdicts.

Updated Apr 26, 2026
Search Tags:#software architecture#implementation plan#design review#bias mitigation#adversarial review#red teaming#verdict gating
One-click install
npx skills add https://github.com/Jin9/skillify-foundation --skill red-teaming-implementation-plan-jin9
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: red-teaming-implementation-plan
Source: https://github.com/Jin9/skillify-foundation/tree/main/treasury/red-teaming-implementation-plan
Command: npx skills add https://github.com/Jin9/skillify-foundation --skill red-teaming-implementation-plan-jin9

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Design flaws discovered after code is written are expensive to fix. This Skill stress-tests a Tech-Lead design or implementation plan before any code is written, catching requirement gaps, architectural weaknesses, and contract risks while they are still cheap to correct. ## Core Features & Use Cases - Adversarial Plan Review: Sweeps eight attack lenses (requirements, architecture, contract, data, security, operability, cost, testability) against the plan and its business brief. - Machine-Readable Verdicts: Emits a structured PROCEED, REVISE, or BLOCK verdict with severity-ranked findings, confidence score, and audit ID that a workflow engine can gate on. - Bias Mitigation: Requires a steelman of the plan first, enforces reviewer-author heterogeneity, and filters hallucinated or unfalsifiable findings. - Use Case: Before fanning out an epic to per-component implementation tasks, run this Skill on the Tech-Lead design to decide whether the plan is safe to build, needs revision, or must be blocked. ## Quick Start Red-team this implementation plan against the attached epic brief and tell me whether it is safe to fan out to per-component work.

Frequently Asked Questions about red-teaming-implementation-plan

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I review an implementation plan before writing code?▼

Submit the plan and optionally the business brief it must satisfy. The skill steelmans the plan, sweeps eight attack lenses covering requirements through testability, then emits a PROCEED, REVISE, or BLOCK verdict with severity-ranked findings.

What is the difference between red-teaming a plan and reviewing a code diff?▼

Plan red-teaming happens before any code exists and critiques the design itself. This skill explicitly does not review code diffs or pull requests; use a code review or pr-design-review tool for that stage.

Can the same AI model that wrote the plan also review it?▼

It can, but the skill records the conflict in bias_checks and lowers the confidence score. A different or stronger reviewer model is preferred because heterogeneous critics reliably outperform self-critique.

When should I not use adversarial plan review?▼

Do not use it for deep security threat-modeling, which belongs to a dedicated security review skill, or to author and revise the plan itself. It is read-only and never edits the plan, code, or fixes.

What happens when the plan input is empty or unparseable?▼

The skill stops without issuing a verdict and returns a needs-input note that escalates to a human queue. A defensible verdict is never forced from missing or malformed input.

Related Skills

Explore More Software Engineering Skills→
rakhnianskii9rakhnianskii9

create-stories

Break epics into auditable stories with TR-ID and ADR references.

Community
Advanced
chughtapanchughtapan

stamina

Dispatch artifacts through multiple review skills and aggregate verdicts.

Community
Advanced
chicongstchicongst
2

security-review

Audit trust boundaries, data flows, and secrets handling to produce a structured security review report.

Community
Advanced
LerianStudioLerianStudio
4

ring:dev-validation

Automate Gate 5 validation requiring explicit human approval and auditable evidence collection.

Official
Basic