mcpField
A
ai-feature-product-risk-review

ai-feature-product-risk-review

github.com/sylphxaiTesting & quality145 installsUpdated today
Repository

About ai-feature-product-risk-review

A skill for product and operations teams assessing risks in AI features. It turns a feature question into a risk matrix, evaluation plan, rollout gates, and incident paths.

Key features

  1. 01AI feature boundary and harm analysis
  2. 02Quality, safety, privacy, and cost mapping
  3. 03Risk matrix with owners and metrics
  4. 04Evaluation and rollout gate planning
  5. 05UI safeguards and incident paths

Use cases

  1. 01Reviewing an AI feature before launch
  2. 02Planning evaluations and monitoring
  3. 03Defining safeguards for agentic actions

Install

Copy the command and run it in a terminal.

$npx skills add https://github.com/sylphxai/skills --skill ai-feature-product-risk-review

After installing, type/ai-feature-product-risk-reviewin Claude Code to confirm it loaded.

Bundled filesagents/openai.yamlreferences/ai-feature-risk-patterns.md

Details

From the project README

AI Feature Product Risk Review

Use this skill to convert a risky product, operations, trust, or marketplace question into a concrete artifact with owners, gates, metrics, and recovery paths.

Workflow

  1. Identify AI job, user harm if wrong, model/input/output boundary, autonomy level, data sensitivity, and business metric.
  2. Read references/ai-feature-risk-patterns.md.
  3. Map quality, safety, privacy, cost, latency, abuse, support, transparency, and fallback requirements.
  4. Separate assistive suggestions, generated content, classification, ranking, automation, and agentic actions.
  5. Produce risk matrix, eval plan, UI guardrails, rollout gates, and support/incident paths.

When not to use

  • Do not use for generic advice the base model already handles without this skill's specific artifact contract.

Guardrails

  • Do not hide uncertainty or present generated output as verified fact when it is not.
  • Do not automate irreversible, financial, legal, safety, account, or privacy-impacting actions without appropriate gates.
  • Do not use user/private data beyond stated purpose, retention, and consent boundaries.
  • AI feature launch needs evals, monitoring, fallback, and support playbooks before broad rollout.

Output format

AI feature context:
Job / autonomy / data / harm:

Risk matrix:
| Risk | Failure mode | User impact | Control | Eval/metric | Owner |
| --- | --- | --- | --- | --- | --- |

UI and rollout gates:
- <transparency/fallback/review/flag>

Support/incident path:
- <case type> -> <triage/evidence/recovery>

Comments

Markdown supported