Add 500 AI skill markdown files

README.md

@@ -0,0 +1,24 @@
+---
+license: mit
+tags:
+- codex-skills
+- ai
+- machine-learning
+- llm
+- web
+- dataset
+pretty_name: AI Skill MD Dataset
+---
+
+# AI Skill MD Dataset
+
+This Hugging Face Dataset repository contains 500 Codex-style skill folders.
+Each skill has a required `SKILL.md` file and lightweight `agents/openai.yaml` metadata.
+
+The repository is intended as a file database of skill definitions only. It is not a Hugging Face Space and contains no app runtime.
+
+## Layout
+
+- `skills/<skill-name>/SKILL.md`: skill instructions and trigger metadata
+- `skills/<skill-name>/agents/openai.yaml`: UI-facing metadata
+- `skills_index.jsonl`: searchable index of all generated skills

skills/a-b-testing-models/SKILL.md

@@ -0,0 +1,29 @@
+---
+name: a-b-testing-models
+description: "Guidance for A B testing models in MLOps. Use when Codex needs to plan, build, review, test, debug, or document work involving A B testing models, mlops, AI systems, software delivery, data workflows, or model quality."
+---
+
+# A B Testing Models
+
+## Core Workflow
+
+1. Confirm the user goal, target environment, inputs, outputs, constraints, and success criteria for A B testing models.
+2. Inspect existing code, data, prompts, models, configs, or product flows before proposing changes.
+3. Choose the smallest implementation path that fits the current system and keeps future maintenance clear.
+4. Add focused validation for behavior, quality, security, privacy, cost, and operational risk where relevant.
+5. Report the concrete change, verification performed, and remaining assumptions.
+
+## Subject Checklist
+
+- Treat this as a MLOps task centered on A B testing models.
+- Version datasets, code, configs, and model artifacts together.
+- Define deploy, monitor, rollback, and incident paths.
+- Automate repeatable training and serving checks.
+- Prefer existing project conventions, libraries, schemas, and deployment patterns.
+- Avoid broad rewrites unless the current structure blocks a correct solution.
+
+## Deliverables
+
+- Produce implementation steps, code edits, tests, prompts, model choices, data checks, or review findings as the task requires.
+- Make tradeoffs explicit: accuracy, latency, cost, safety, maintainability, and user experience.
+- Leave enough context for another engineer or agent to reproduce the result.

skills/a-b-testing-models/agents/openai.yaml

@@ -0,0 +1,3 @@
+display_name: "A B Testing Models"
+short_description: "Work on A B testing models for MLOps."
+default_prompt: "Use this skill to help with A B testing models in MLOps."

skills/ablation-studies/SKILL.md

@@ -0,0 +1,29 @@
+---
+name: ablation-studies
+description: "Guidance for ablation studies in Research And Scientific AI. Use when Codex needs to plan, build, review, test, debug, or document work involving ablation studies, research and scientific ai, AI systems, software delivery, data workflows, or model quality."
+---
+
+# Ablation Studies
+
+## Core Workflow
+
+1. Confirm the user goal, target environment, inputs, outputs, constraints, and success criteria for ablation studies.
+2. Inspect existing code, data, prompts, models, configs, or product flows before proposing changes.
+3. Choose the smallest implementation path that fits the current system and keeps future maintenance clear.
+4. Add focused validation for behavior, quality, security, privacy, cost, and operational risk where relevant.
+5. Report the concrete change, verification performed, and remaining assumptions.
+
+## Subject Checklist
+
+- Treat this as a Research And Scientific AI task centered on ablation studies.
+- Preserve provenance, seeds, configs, and environment details.
+- Separate claims, experiments, and evidence.
+- Package artifacts so another researcher can reproduce results.
+- Prefer existing project conventions, libraries, schemas, and deployment patterns.
+- Avoid broad rewrites unless the current structure blocks a correct solution.
+
+## Deliverables
+
+- Produce implementation steps, code edits, tests, prompts, model choices, data checks, or review findings as the task requires.
+- Make tradeoffs explicit: accuracy, latency, cost, safety, maintainability, and user experience.
+- Leave enough context for another engineer or agent to reproduce the result.

skills/ablation-studies/agents/openai.yaml

@@ -0,0 +1,3 @@
+display_name: "Ablation Studies"
+short_description: "Work on ablation studies for Research And Scientific AI."
+default_prompt: "Use this skill to help with ablation studies in Research And Scientific AI."

skills/accent-robustness/SKILL.md

@@ -0,0 +1,29 @@
+---
+name: accent-robustness
+description: "Guidance for accent robustness in Speech And Audio. Use when Codex needs to plan, build, review, test, debug, or document work involving accent robustness, speech and audio, AI systems, software delivery, data workflows, or model quality."
+---
+
+# Accent Robustness
+
+## Core Workflow
+
+1. Confirm the user goal, target environment, inputs, outputs, constraints, and success criteria for accent robustness.
+2. Inspect existing code, data, prompts, models, configs, or product flows before proposing changes.
+3. Choose the smallest implementation path that fits the current system and keeps future maintenance clear.
+4. Add focused validation for behavior, quality, security, privacy, cost, and operational risk where relevant.
+5. Report the concrete change, verification performed, and remaining assumptions.
+
+## Subject Checklist
+
+- Treat this as a Speech And Audio task centered on accent robustness.
+- Confirm sample rate, channels, codec, and segmentation assumptions.
+- Evaluate latency and word level quality on real audio.
+- Handle noisy, accented, and interrupted speech explicitly.
+- Prefer existing project conventions, libraries, schemas, and deployment patterns.
+- Avoid broad rewrites unless the current structure blocks a correct solution.
+
+## Deliverables
+
+- Produce implementation steps, code edits, tests, prompts, model choices, data checks, or review findings as the task requires.
+- Make tradeoffs explicit: accuracy, latency, cost, safety, maintainability, and user experience.
+- Leave enough context for another engineer or agent to reproduce the result.

skills/accent-robustness/agents/openai.yaml

@@ -0,0 +1,3 @@
+display_name: "Accent Robustness"
+short_description: "Work on accent robustness for Speech And Audio."
+default_prompt: "Use this skill to help with accent robustness in Speech And Audio."

skills/accessibility-for-ai-products/SKILL.md

@@ -0,0 +1,29 @@
+---
+name: accessibility-for-ai-products
+description: "Guidance for accessibility for AI products in AI Product And UX. Use when Codex needs to plan, build, review, test, debug, or document work involving accessibility for AI products, ai product and ux, AI systems, software delivery, data workflows, or model quality."
+---
+
+# Accessibility For AI Products
+
+## Core Workflow
+
+1. Confirm the user goal, target environment, inputs, outputs, constraints, and success criteria for accessibility for AI products.
+2. Inspect existing code, data, prompts, models, configs, or product flows before proposing changes.
+3. Choose the smallest implementation path that fits the current system and keeps future maintenance clear.
+4. Add focused validation for behavior, quality, security, privacy, cost, and operational risk where relevant.
+5. Report the concrete change, verification performed, and remaining assumptions.
+
+## Subject Checklist
+
+- Treat this as a AI Product And UX task centered on accessibility for AI products.
+- Start from user job, risk, and feedback loop.
+- Expose uncertainty and recovery paths without clutter.
+- Measure usefulness, trust, and operational burden.
+- Prefer existing project conventions, libraries, schemas, and deployment patterns.
+- Avoid broad rewrites unless the current structure blocks a correct solution.
+
+## Deliverables
+
+- Produce implementation steps, code edits, tests, prompts, model choices, data checks, or review findings as the task requires.
+- Make tradeoffs explicit: accuracy, latency, cost, safety, maintainability, and user experience.
+- Leave enough context for another engineer or agent to reproduce the result.

skills/accessibility-for-ai-products/agents/openai.yaml

@@ -0,0 +1,3 @@
+display_name: "Accessibility For AI Products"
+short_description: "Work on accessibility for AI products for AI Product And UX."
+default_prompt: "Use this skill to help with accessibility for AI products in AI Product And UX."

skills/action-recognition/SKILL.md

@@ -0,0 +1,29 @@
+---
+name: action-recognition
+description: "Guidance for action recognition in Computer Vision. Use when Codex needs to plan, build, review, test, debug, or document work involving action recognition, computer vision, AI systems, software delivery, data workflows, or model quality."
+---
+
+# Action Recognition
+
+## Core Workflow
+
+1. Confirm the user goal, target environment, inputs, outputs, constraints, and success criteria for action recognition.
+2. Inspect existing code, data, prompts, models, configs, or product flows before proposing changes.
+3. Choose the smallest implementation path that fits the current system and keeps future maintenance clear.
+4. Add focused validation for behavior, quality, security, privacy, cost, and operational risk where relevant.
+5. Report the concrete change, verification performed, and remaining assumptions.
+
+## Subject Checklist
+
+- Treat this as a Computer Vision task centered on action recognition.
+- Inspect images visually before trusting labels.
+- Separate augmentation, preprocessing, and model inference concerns.
+- Measure performance across lighting, scale, occlusion, and device sources.
+- Prefer existing project conventions, libraries, schemas, and deployment patterns.
+- Avoid broad rewrites unless the current structure blocks a correct solution.
+
+## Deliverables
+
+- Produce implementation steps, code edits, tests, prompts, model choices, data checks, or review findings as the task requires.
+- Make tradeoffs explicit: accuracy, latency, cost, safety, maintainability, and user experience.
+- Leave enough context for another engineer or agent to reproduce the result.

skills/action-recognition/agents/openai.yaml

@@ -0,0 +1,3 @@
+display_name: "Action Recognition"
+short_description: "Work on action recognition for Computer Vision."
+default_prompt: "Use this skill to help with action recognition in Computer Vision."

skills/active-learning/SKILL.md

@@ -0,0 +1,29 @@
+---
+name: active-learning
+description: "Guidance for active learning in Machine Learning. Use when Codex needs to plan, build, review, test, debug, or document work involving active learning, machine learning, AI systems, software delivery, data workflows, or model quality."
+---
+
+# Active Learning
+
+## Core Workflow
+
+1. Confirm the user goal, target environment, inputs, outputs, constraints, and success criteria for active learning.
+2. Inspect existing code, data, prompts, models, configs, or product flows before proposing changes.
+3. Choose the smallest implementation path that fits the current system and keeps future maintenance clear.
+4. Add focused validation for behavior, quality, security, privacy, cost, and operational risk where relevant.
+5. Report the concrete change, verification performed, and remaining assumptions.
+
+## Subject Checklist
+
+- Treat this as a Machine Learning task centered on active learning.
+- Start with a simple baseline and leakage checks.
+- Keep train, validation, and test boundaries explicit.
+- Report metrics that match the product decision.
+- Prefer existing project conventions, libraries, schemas, and deployment patterns.
+- Avoid broad rewrites unless the current structure blocks a correct solution.
+
+## Deliverables
+
+- Produce implementation steps, code edits, tests, prompts, model choices, data checks, or review findings as the task requires.
+- Make tradeoffs explicit: accuracy, latency, cost, safety, maintainability, and user experience.
+- Leave enough context for another engineer or agent to reproduce the result.

skills/active-learning/agents/openai.yaml

@@ -0,0 +1,3 @@
+display_name: "Active Learning"
+short_description: "Work on active learning for Machine Learning."
+default_prompt: "Use this skill to help with active learning in Machine Learning."

skills/admin-dashboards/SKILL.md

@@ -0,0 +1,29 @@
+---
+name: admin-dashboards
+description: "Guidance for admin dashboards in Web Engineering. Use when Codex needs to plan, build, review, test, debug, or document work involving admin dashboards, web engineering, AI systems, software delivery, data workflows, or model quality."
+---
+
+# Admin Dashboards
+
+## Core Workflow
+
+1. Confirm the user goal, target environment, inputs, outputs, constraints, and success criteria for admin dashboards.
+2. Inspect existing code, data, prompts, models, configs, or product flows before proposing changes.
+3. Choose the smallest implementation path that fits the current system and keeps future maintenance clear.
+4. Add focused validation for behavior, quality, security, privacy, cost, and operational risk where relevant.
+5. Report the concrete change, verification performed, and remaining assumptions.
+
+## Subject Checklist
+
+- Treat this as a Web Engineering task centered on admin dashboards.
+- Follow the existing framework, routing, and component patterns.
+- Make loading, empty, error, and mobile states explicit.
+- Verify accessibility and performance before handoff.
+- Prefer existing project conventions, libraries, schemas, and deployment patterns.
+- Avoid broad rewrites unless the current structure blocks a correct solution.
+
+## Deliverables
+
+- Produce implementation steps, code edits, tests, prompts, model choices, data checks, or review findings as the task requires.
+- Make tradeoffs explicit: accuracy, latency, cost, safety, maintainability, and user experience.
+- Leave enough context for another engineer or agent to reproduce the result.

skills/admin-dashboards/agents/openai.yaml

@@ -0,0 +1,3 @@
+display_name: "Admin Dashboards"
+short_description: "Work on admin dashboards for Web Engineering."
+default_prompt: "Use this skill to help with admin dashboards in Web Engineering."

skills/adversarial-prompt-tests/SKILL.md

@@ -0,0 +1,29 @@
+---
+name: adversarial-prompt-tests
+description: "Guidance for adversarial prompt tests in Prompting And Evaluation. Use when Codex needs to plan, build, review, test, debug, or document work involving adversarial prompt tests, prompting and evaluation, AI systems, software delivery, data workflows, or model quality."
+---
+
+# Adversarial Prompt Tests
+
+## Core Workflow
+
+1. Confirm the user goal, target environment, inputs, outputs, constraints, and success criteria for adversarial prompt tests.
+2. Inspect existing code, data, prompts, models, configs, or product flows before proposing changes.
+3. Choose the smallest implementation path that fits the current system and keeps future maintenance clear.
+4. Add focused validation for behavior, quality, security, privacy, cost, and operational risk where relevant.
+5. Report the concrete change, verification performed, and remaining assumptions.
+
+## Subject Checklist
+
+- Treat this as a Prompting And Evaluation task centered on adversarial prompt tests.
+- Create representative positive, negative, and adversarial examples.
+- Separate prompt changes from model changes during evaluation.
+- Use measurable rubrics instead of vague quality language.
+- Prefer existing project conventions, libraries, schemas, and deployment patterns.
+- Avoid broad rewrites unless the current structure blocks a correct solution.
+
+## Deliverables
+
+- Produce implementation steps, code edits, tests, prompts, model choices, data checks, or review findings as the task requires.
+- Make tradeoffs explicit: accuracy, latency, cost, safety, maintainability, and user experience.
+- Leave enough context for another engineer or agent to reproduce the result.

skills/adversarial-prompt-tests/agents/openai.yaml

@@ -0,0 +1,3 @@
+display_name: "Adversarial Prompt Tests"
+short_description: "Work on adversarial prompt tests for Prompting And Evaluation."
+default_prompt: "Use this skill to help with adversarial prompt tests in Prompting And Evaluation."

skills/agent-evaluation-harnesses/SKILL.md

@@ -0,0 +1,29 @@
+---
+name: agent-evaluation-harnesses
+description: "Guidance for agent evaluation harnesses in Agentic AI. Use when Codex needs to plan, build, review, test, debug, or document work involving agent evaluation harnesses, agentic ai, AI systems, software delivery, data workflows, or model quality."
+---
+
+# Agent Evaluation Harnesses
+
+## Core Workflow
+
+1. Confirm the user goal, target environment, inputs, outputs, constraints, and success criteria for agent evaluation harnesses.
+2. Inspect existing code, data, prompts, models, configs, or product flows before proposing changes.
+3. Choose the smallest implementation path that fits the current system and keeps future maintenance clear.
+4. Add focused validation for behavior, quality, security, privacy, cost, and operational risk where relevant.
+5. Report the concrete change, verification performed, and remaining assumptions.
+
+## Subject Checklist
+
+- Treat this as a Agentic AI task centered on agent evaluation harnesses.
+- Model the workflow state explicitly.
+- Gate risky actions behind user approval or dry runs.
+- Log each tool call, observation, and state transition for review.
+- Prefer existing project conventions, libraries, schemas, and deployment patterns.
+- Avoid broad rewrites unless the current structure blocks a correct solution.
+
+## Deliverables
+
+- Produce implementation steps, code edits, tests, prompts, model choices, data checks, or review findings as the task requires.
+- Make tradeoffs explicit: accuracy, latency, cost, safety, maintainability, and user experience.
+- Leave enough context for another engineer or agent to reproduce the result.

skills/agent-evaluation-harnesses/agents/openai.yaml

@@ -0,0 +1,3 @@
+display_name: "Agent Evaluation Harnesses"
+short_description: "Work on agent evaluation harnesses for Agentic AI."
+default_prompt: "Use this skill to help with agent evaluation harnesses in Agentic AI."

skills/agent-memory-stores/SKILL.md

@@ -0,0 +1,29 @@
+---
+name: agent-memory-stores
+description: "Guidance for agent memory stores in Agentic AI. Use when Codex needs to plan, build, review, test, debug, or document work involving agent memory stores, agentic ai, AI systems, software delivery, data workflows, or model quality."
+---
+
+# Agent Memory Stores
+
+## Core Workflow
+
+1. Confirm the user goal, target environment, inputs, outputs, constraints, and success criteria for agent memory stores.
+2. Inspect existing code, data, prompts, models, configs, or product flows before proposing changes.
+3. Choose the smallest implementation path that fits the current system and keeps future maintenance clear.
+4. Add focused validation for behavior, quality, security, privacy, cost, and operational risk where relevant.
+5. Report the concrete change, verification performed, and remaining assumptions.
+
+## Subject Checklist
+
+- Treat this as a Agentic AI task centered on agent memory stores.
+- Model the workflow state explicitly.
+- Gate risky actions behind user approval or dry runs.
+- Log each tool call, observation, and state transition for review.
+- Prefer existing project conventions, libraries, schemas, and deployment patterns.
+- Avoid broad rewrites unless the current structure blocks a correct solution.
+
+## Deliverables
+
+- Produce implementation steps, code edits, tests, prompts, model choices, data checks, or review findings as the task requires.
+- Make tradeoffs explicit: accuracy, latency, cost, safety, maintainability, and user experience.
+- Leave enough context for another engineer or agent to reproduce the result.

skills/agent-memory-stores/agents/openai.yaml

@@ -0,0 +1,3 @@
+display_name: "Agent Memory Stores"
+short_description: "Work on agent memory stores for Agentic AI."
+default_prompt: "Use this skill to help with agent memory stores in Agentic AI."

skills/agent-trace-analysis/SKILL.md

@@ -0,0 +1,29 @@
+---
+name: agent-trace-analysis
+description: "Guidance for agent trace analysis in Agentic AI. Use when Codex needs to plan, build, review, test, debug, or document work involving agent trace analysis, agentic ai, AI systems, software delivery, data workflows, or model quality."
+---
+
+# Agent Trace Analysis
+
+## Core Workflow
+
+1. Confirm the user goal, target environment, inputs, outputs, constraints, and success criteria for agent trace analysis.
+2. Inspect existing code, data, prompts, models, configs, or product flows before proposing changes.
+3. Choose the smallest implementation path that fits the current system and keeps future maintenance clear.
+4. Add focused validation for behavior, quality, security, privacy, cost, and operational risk where relevant.
+5. Report the concrete change, verification performed, and remaining assumptions.
+
+## Subject Checklist
+
+- Treat this as a Agentic AI task centered on agent trace analysis.
+- Model the workflow state explicitly.
+- Gate risky actions behind user approval or dry runs.
+- Log each tool call, observation, and state transition for review.
+- Prefer existing project conventions, libraries, schemas, and deployment patterns.
+- Avoid broad rewrites unless the current structure blocks a correct solution.
+
+## Deliverables
+
+- Produce implementation steps, code edits, tests, prompts, model choices, data checks, or review findings as the task requires.
+- Make tradeoffs explicit: accuracy, latency, cost, safety, maintainability, and user experience.
+- Leave enough context for another engineer or agent to reproduce the result.

skills/agent-trace-analysis/agents/openai.yaml

@@ -0,0 +1,3 @@
+display_name: "Agent Trace Analysis"
+short_description: "Work on agent trace analysis for Agentic AI."
+default_prompt: "Use this skill to help with agent trace analysis in Agentic AI."

skills/agriculture-ai-workflows/SKILL.md

@@ -0,0 +1,29 @@
+---
+name: agriculture-ai-workflows
+description: "Guidance for agriculture AI workflows in Domain AI. Use when Codex needs to plan, build, review, test, debug, or document work involving agriculture AI workflows, domain ai, AI systems, software delivery, data workflows, or model quality."
+---
+
+# Agriculture AI Workflows
+
+## Core Workflow
+
+1. Confirm the user goal, target environment, inputs, outputs, constraints, and success criteria for agriculture AI workflows.
+2. Inspect existing code, data, prompts, models, configs, or product flows before proposing changes.
+3. Choose the smallest implementation path that fits the current system and keeps future maintenance clear.
+4. Add focused validation for behavior, quality, security, privacy, cost, and operational risk where relevant.
+5. Report the concrete change, verification performed, and remaining assumptions.
+
+## Subject Checklist
+
+- Treat this as a Domain AI task centered on agriculture AI workflows.
+- Confirm domain regulations, workflow owners, and data sensitivity.
+- Keep humans accountable for high impact decisions.
+- Validate outputs with domain specific examples and review paths.
+- Prefer existing project conventions, libraries, schemas, and deployment patterns.
+- Avoid broad rewrites unless the current structure blocks a correct solution.
+
+## Deliverables
+
+- Produce implementation steps, code edits, tests, prompts, model choices, data checks, or review findings as the task requires.
+- Make tradeoffs explicit: accuracy, latency, cost, safety, maintainability, and user experience.
+- Leave enough context for another engineer or agent to reproduce the result.

skills/agriculture-ai-workflows/agents/openai.yaml

@@ -0,0 +1,3 @@
+display_name: "Agriculture AI Workflows"
+short_description: "Work on agriculture AI workflows for Domain AI."
+default_prompt: "Use this skill to help with agriculture AI workflows in Domain AI."

skills/ai-admin-controls/SKILL.md

@@ -0,0 +1,29 @@
+---
+name: ai-admin-controls
+description: "Guidance for AI admin controls in AI Product And UX. Use when Codex needs to plan, build, review, test, debug, or document work involving AI admin controls, ai product and ux, AI systems, software delivery, data workflows, or model quality."
+---
+
+# AI Admin Controls
+
+## Core Workflow
+
+1. Confirm the user goal, target environment, inputs, outputs, constraints, and success criteria for AI admin controls.
+2. Inspect existing code, data, prompts, models, configs, or product flows before proposing changes.
+3. Choose the smallest implementation path that fits the current system and keeps future maintenance clear.
+4. Add focused validation for behavior, quality, security, privacy, cost, and operational risk where relevant.
+5. Report the concrete change, verification performed, and remaining assumptions.
+
+## Subject Checklist
+
+- Treat this as a AI Product And UX task centered on AI admin controls.
+- Start from user job, risk, and feedback loop.
+- Expose uncertainty and recovery paths without clutter.
+- Measure usefulness, trust, and operational burden.
+- Prefer existing project conventions, libraries, schemas, and deployment patterns.
+- Avoid broad rewrites unless the current structure blocks a correct solution.
+
+## Deliverables
+
+- Produce implementation steps, code edits, tests, prompts, model choices, data checks, or review findings as the task requires.
+- Make tradeoffs explicit: accuracy, latency, cost, safety, maintainability, and user experience.
+- Leave enough context for another engineer or agent to reproduce the result.

skills/ai-admin-controls/agents/openai.yaml

@@ -0,0 +1,3 @@
+display_name: "AI Admin Controls"
+short_description: "Work on AI admin controls for AI Product And UX."
+default_prompt: "Use this skill to help with AI admin controls in AI Product And UX."

skills/ai-collaboration-features/SKILL.md

@@ -0,0 +1,29 @@
+---
+name: ai-collaboration-features
+description: "Guidance for AI collaboration features in AI Product And UX. Use when Codex needs to plan, build, review, test, debug, or document work involving AI collaboration features, ai product and ux, AI systems, software delivery, data workflows, or model quality."
+---
+
+# AI Collaboration Features
+
+## Core Workflow
+
+1. Confirm the user goal, target environment, inputs, outputs, constraints, and success criteria for AI collaboration features.
+2. Inspect existing code, data, prompts, models, configs, or product flows before proposing changes.
+3. Choose the smallest implementation path that fits the current system and keeps future maintenance clear.
+4. Add focused validation for behavior, quality, security, privacy, cost, and operational risk where relevant.
+5. Report the concrete change, verification performed, and remaining assumptions.
+
+## Subject Checklist
+
+- Treat this as a AI Product And UX task centered on AI collaboration features.
+- Start from user job, risk, and feedback loop.
+- Expose uncertainty and recovery paths without clutter.
+- Measure usefulness, trust, and operational burden.
+- Prefer existing project conventions, libraries, schemas, and deployment patterns.
+- Avoid broad rewrites unless the current structure blocks a correct solution.
+
+## Deliverables
+
+- Produce implementation steps, code edits, tests, prompts, model choices, data checks, or review findings as the task requires.
+- Make tradeoffs explicit: accuracy, latency, cost, safety, maintainability, and user experience.
+- Leave enough context for another engineer or agent to reproduce the result.

skills/ai-collaboration-features/agents/openai.yaml

@@ -0,0 +1,3 @@
+display_name: "AI Collaboration Features"
+short_description: "Work on AI collaboration features for AI Product And UX."
+default_prompt: "Use this skill to help with AI collaboration features in AI Product And UX."

skills/ai-disclosure-copy/SKILL.md

@@ -0,0 +1,29 @@
+---
+name: ai-disclosure-copy
+description: "Guidance for AI disclosure copy in AI Product And UX. Use when Codex needs to plan, build, review, test, debug, or document work involving AI disclosure copy, ai product and ux, AI systems, software delivery, data workflows, or model quality."
+---
+
+# AI Disclosure Copy
+
+## Core Workflow
+
+1. Confirm the user goal, target environment, inputs, outputs, constraints, and success criteria for AI disclosure copy.
+2. Inspect existing code, data, prompts, models, configs, or product flows before proposing changes.
+3. Choose the smallest implementation path that fits the current system and keeps future maintenance clear.
+4. Add focused validation for behavior, quality, security, privacy, cost, and operational risk where relevant.
+5. Report the concrete change, verification performed, and remaining assumptions.
+
+## Subject Checklist
+
+- Treat this as a AI Product And UX task centered on AI disclosure copy.
+- Start from user job, risk, and feedback loop.
+- Expose uncertainty and recovery paths without clutter.
+- Measure usefulness, trust, and operational burden.
+- Prefer existing project conventions, libraries, schemas, and deployment patterns.
+- Avoid broad rewrites unless the current structure blocks a correct solution.
+
+## Deliverables
+
+- Produce implementation steps, code edits, tests, prompts, model choices, data checks, or review findings as the task requires.
+- Make tradeoffs explicit: accuracy, latency, cost, safety, maintainability, and user experience.
+- Leave enough context for another engineer or agent to reproduce the result.

skills/ai-disclosure-copy/agents/openai.yaml

@@ -0,0 +1,3 @@
+display_name: "AI Disclosure Copy"
+short_description: "Work on AI disclosure copy for AI Product And UX."
+default_prompt: "Use this skill to help with AI disclosure copy in AI Product And UX."

skills/ai-feature-scoping/SKILL.md

@@ -0,0 +1,29 @@
+---
+name: ai-feature-scoping
+description: "Guidance for AI feature scoping in AI Product And UX. Use when Codex needs to plan, build, review, test, debug, or document work involving AI feature scoping, ai product and ux, AI systems, software delivery, data workflows, or model quality."
+---
+
+# AI Feature Scoping
+
+## Core Workflow
+
+1. Confirm the user goal, target environment, inputs, outputs, constraints, and success criteria for AI feature scoping.
+2. Inspect existing code, data, prompts, models, configs, or product flows before proposing changes.
+3. Choose the smallest implementation path that fits the current system and keeps future maintenance clear.
+4. Add focused validation for behavior, quality, security, privacy, cost, and operational risk where relevant.
+5. Report the concrete change, verification performed, and remaining assumptions.
+
+## Subject Checklist
+
+- Treat this as a AI Product And UX task centered on AI feature scoping.
+- Start from user job, risk, and feedback loop.
+- Expose uncertainty and recovery paths without clutter.
+- Measure usefulness, trust, and operational burden.
+- Prefer existing project conventions, libraries, schemas, and deployment patterns.
+- Avoid broad rewrites unless the current structure blocks a correct solution.
+
+## Deliverables
+
+- Produce implementation steps, code edits, tests, prompts, model choices, data checks, or review findings as the task requires.
+- Make tradeoffs explicit: accuracy, latency, cost, safety, maintainability, and user experience.
+- Leave enough context for another engineer or agent to reproduce the result.

skills/ai-feature-scoping/agents/openai.yaml

@@ -0,0 +1,3 @@
+display_name: "AI Feature Scoping"
+short_description: "Work on AI feature scoping for AI Product And UX."
+default_prompt: "Use this skill to help with AI feature scoping in AI Product And UX."

skills/ai-productivity-tools/SKILL.md

@@ -0,0 +1,29 @@
+---
+name: ai-productivity-tools
+description: "Guidance for AI productivity tools in AI Product And UX. Use when Codex needs to plan, build, review, test, debug, or document work involving AI productivity tools, ai product and ux, AI systems, software delivery, data workflows, or model quality."
+---
+
+# AI Productivity Tools
+
+## Core Workflow
+
+1. Confirm the user goal, target environment, inputs, outputs, constraints, and success criteria for AI productivity tools.
+2. Inspect existing code, data, prompts, models, configs, or product flows before proposing changes.
+3. Choose the smallest implementation path that fits the current system and keeps future maintenance clear.
+4. Add focused validation for behavior, quality, security, privacy, cost, and operational risk where relevant.
+5. Report the concrete change, verification performed, and remaining assumptions.
+
+## Subject Checklist
+
+- Treat this as a AI Product And UX task centered on AI productivity tools.
+- Start from user job, risk, and feedback loop.
+- Expose uncertainty and recovery paths without clutter.
+- Measure usefulness, trust, and operational burden.
+- Prefer existing project conventions, libraries, schemas, and deployment patterns.
+- Avoid broad rewrites unless the current structure blocks a correct solution.
+
+## Deliverables
+
+- Produce implementation steps, code edits, tests, prompts, model choices, data checks, or review findings as the task requires.
+- Make tradeoffs explicit: accuracy, latency, cost, safety, maintainability, and user experience.
+- Leave enough context for another engineer or agent to reproduce the result.

skills/ai-productivity-tools/agents/openai.yaml

@@ -0,0 +1,3 @@
+display_name: "AI Productivity Tools"
+short_description: "Work on AI productivity tools for AI Product And UX."
+default_prompt: "Use this skill to help with AI productivity tools in AI Product And UX."

skills/ai-safety-policy/SKILL.md

@@ -0,0 +1,29 @@
+---
+name: ai-safety-policy
+description: "Guidance for AI safety policy in Security And Privacy. Use when Codex needs to plan, build, review, test, debug, or document work involving AI safety policy, security and privacy, AI systems, software delivery, data workflows, or model quality."
+---
+
+# AI Safety Policy
+
+## Core Workflow
+
+1. Confirm the user goal, target environment, inputs, outputs, constraints, and success criteria for AI safety policy.
+2. Inspect existing code, data, prompts, models, configs, or product flows before proposing changes.
+3. Choose the smallest implementation path that fits the current system and keeps future maintenance clear.
+4. Add focused validation for behavior, quality, security, privacy, cost, and operational risk where relevant.
+5. Report the concrete change, verification performed, and remaining assumptions.
+
+## Subject Checklist
+
+- Treat this as a Security And Privacy task centered on AI safety policy.
+- Identify assets, trust boundaries, and abuse cases first.
+- Minimize sensitive data collection and retention.
+- Verify controls with tests, logs, and reviewable evidence.
+- Prefer existing project conventions, libraries, schemas, and deployment patterns.
+- Avoid broad rewrites unless the current structure blocks a correct solution.
+
+## Deliverables
+
+- Produce implementation steps, code edits, tests, prompts, model choices, data checks, or review findings as the task requires.
+- Make tradeoffs explicit: accuracy, latency, cost, safety, maintainability, and user experience.
+- Leave enough context for another engineer or agent to reproduce the result.

skills/ai-safety-policy/agents/openai.yaml

@@ -0,0 +1,3 @@
+display_name: "AI Safety Policy"
+short_description: "Work on AI safety policy for Security And Privacy."
+default_prompt: "Use this skill to help with AI safety policy in Security And Privacy."

skills/ai-settings-panels/SKILL.md

@@ -0,0 +1,29 @@
+---
+name: ai-settings-panels
+description: "Guidance for AI settings panels in AI Product And UX. Use when Codex needs to plan, build, review, test, debug, or document work involving AI settings panels, ai product and ux, AI systems, software delivery, data workflows, or model quality."
+---
+
+# AI Settings Panels
+
+## Core Workflow
+
+1. Confirm the user goal, target environment, inputs, outputs, constraints, and success criteria for AI settings panels.
+2. Inspect existing code, data, prompts, models, configs, or product flows before proposing changes.
+3. Choose the smallest implementation path that fits the current system and keeps future maintenance clear.
+4. Add focused validation for behavior, quality, security, privacy, cost, and operational risk where relevant.
+5. Report the concrete change, verification performed, and remaining assumptions.
+
+## Subject Checklist
+
+- Treat this as a AI Product And UX task centered on AI settings panels.
+- Start from user job, risk, and feedback loop.
+- Expose uncertainty and recovery paths without clutter.
+- Measure usefulness, trust, and operational burden.
+- Prefer existing project conventions, libraries, schemas, and deployment patterns.
+- Avoid broad rewrites unless the current structure blocks a correct solution.
+
+## Deliverables
+
+- Produce implementation steps, code edits, tests, prompts, model choices, data checks, or review findings as the task requires.
+- Make tradeoffs explicit: accuracy, latency, cost, safety, maintainability, and user experience.
+- Leave enough context for another engineer or agent to reproduce the result.

skills/ai-settings-panels/agents/openai.yaml

@@ -0,0 +1,3 @@
+display_name: "AI Settings Panels"
+short_description: "Work on AI settings panels for AI Product And UX."
+default_prompt: "Use this skill to help with AI settings panels in AI Product And UX."

skills/ai-user-onboarding/SKILL.md

@@ -0,0 +1,29 @@
+---
+name: ai-user-onboarding
+description: "Guidance for AI user onboarding in AI Product And UX. Use when Codex needs to plan, build, review, test, debug, or document work involving AI user onboarding, ai product and ux, AI systems, software delivery, data workflows, or model quality."
+---
+
+# AI User Onboarding
+
+## Core Workflow
+
+1. Confirm the user goal, target environment, inputs, outputs, constraints, and success criteria for AI user onboarding.
+2. Inspect existing code, data, prompts, models, configs, or product flows before proposing changes.
+3. Choose the smallest implementation path that fits the current system and keeps future maintenance clear.
+4. Add focused validation for behavior, quality, security, privacy, cost, and operational risk where relevant.
+5. Report the concrete change, verification performed, and remaining assumptions.
+
+## Subject Checklist
+
+- Treat this as a AI Product And UX task centered on AI user onboarding.
+- Start from user job, risk, and feedback loop.
+- Expose uncertainty and recovery paths without clutter.
+- Measure usefulness, trust, and operational burden.
+- Prefer existing project conventions, libraries, schemas, and deployment patterns.
+- Avoid broad rewrites unless the current structure blocks a correct solution.
+
+## Deliverables
+
+- Produce implementation steps, code edits, tests, prompts, model choices, data checks, or review findings as the task requires.
+- Make tradeoffs explicit: accuracy, latency, cost, safety, maintainability, and user experience.
+- Leave enough context for another engineer or agent to reproduce the result.

skills/ai-user-onboarding/agents/openai.yaml

@@ -0,0 +1,3 @@
+display_name: "AI User Onboarding"
+short_description: "Work on AI user onboarding for AI Product And UX."
+default_prompt: "Use this skill to help with AI user onboarding in AI Product And UX."

skills/analytics-event-schemas/SKILL.md

@@ -0,0 +1,29 @@
+---
+name: analytics-event-schemas
+description: "Guidance for analytics event schemas in Databases And Analytics. Use when Codex needs to plan, build, review, test, debug, or document work involving analytics event schemas, databases and analytics, AI systems, software delivery, data workflows, or model quality."
+---
+
+# Analytics Event Schemas
+
+## Core Workflow
+
+1. Confirm the user goal, target environment, inputs, outputs, constraints, and success criteria for analytics event schemas.
+2. Inspect existing code, data, prompts, models, configs, or product flows before proposing changes.
+3. Choose the smallest implementation path that fits the current system and keeps future maintenance clear.
+4. Add focused validation for behavior, quality, security, privacy, cost, and operational risk where relevant.
+5. Report the concrete change, verification performed, and remaining assumptions.
+
+## Subject Checklist
+
+- Treat this as a Databases And Analytics task centered on analytics event schemas.
+- Start from query patterns and data ownership.
+- Validate indexes, constraints, migrations, and rollback plans.
+- Define metrics in one governed layer when possible.
+- Prefer existing project conventions, libraries, schemas, and deployment patterns.
+- Avoid broad rewrites unless the current structure blocks a correct solution.
+
+## Deliverables
+
+- Produce implementation steps, code edits, tests, prompts, model choices, data checks, or review findings as the task requires.
+- Make tradeoffs explicit: accuracy, latency, cost, safety, maintainability, and user experience.
+- Leave enough context for another engineer or agent to reproduce the result.

skills/analytics-event-schemas/agents/openai.yaml

@@ -0,0 +1,3 @@
+display_name: "Analytics Event Schemas"
+short_description: "Work on analytics event schemas for Databases And Analytics."
+default_prompt: "Use this skill to help with analytics event schemas in Databases And Analytics."

skills/android-kotlin-apps/SKILL.md

@@ -0,0 +1,29 @@
+---
+name: android-kotlin-apps
+description: "Guidance for Android Kotlin apps in Mobile App Engineering. Use when Codex needs to plan, build, review, test, debug, or document work involving Android Kotlin apps, mobile app engineering, AI systems, software delivery, data workflows, or model quality."
+---
+
+# Android Kotlin Apps
+
+## Core Workflow
+
+1. Confirm the user goal, target environment, inputs, outputs, constraints, and success criteria for Android Kotlin apps.
+2. Inspect existing code, data, prompts, models, configs, or product flows before proposing changes.
+3. Choose the smallest implementation path that fits the current system and keeps future maintenance clear.
+4. Add focused validation for behavior, quality, security, privacy, cost, and operational risk where relevant.
+5. Report the concrete change, verification performed, and remaining assumptions.
+
+## Subject Checklist
+
+- Treat this as a Mobile App Engineering task centered on Android Kotlin apps.
+- Design for offline, permissions, device sizes, and release stores.
+- Test on realistic devices or emulators.
+- Protect local data and background tasks deliberately.
+- Prefer existing project conventions, libraries, schemas, and deployment patterns.
+- Avoid broad rewrites unless the current structure blocks a correct solution.
+
+## Deliverables
+
+- Produce implementation steps, code edits, tests, prompts, model choices, data checks, or review findings as the task requires.
+- Make tradeoffs explicit: accuracy, latency, cost, safety, maintainability, and user experience.
+- Leave enough context for another engineer or agent to reproduce the result.

skills/android-kotlin-apps/agents/openai.yaml

@@ -0,0 +1,3 @@
+display_name: "Android Kotlin Apps"
+short_description: "Work on Android Kotlin apps for Mobile App Engineering."
+default_prompt: "Use this skill to help with Android Kotlin apps in Mobile App Engineering."

skills/annotation-workflows/SKILL.md

@@ -0,0 +1,29 @@
+---
+name: annotation-workflows
+description: "Guidance for annotation workflows in AI Product And UX. Use when Codex needs to plan, build, review, test, debug, or document work involving annotation workflows, ai product and ux, AI systems, software delivery, data workflows, or model quality."
+---
+
+# Annotation Workflows
+
+## Core Workflow
+
+1. Confirm the user goal, target environment, inputs, outputs, constraints, and success criteria for annotation workflows.
+2. Inspect existing code, data, prompts, models, configs, or product flows before proposing changes.
+3. Choose the smallest implementation path that fits the current system and keeps future maintenance clear.
+4. Add focused validation for behavior, quality, security, privacy, cost, and operational risk where relevant.
+5. Report the concrete change, verification performed, and remaining assumptions.
+
+## Subject Checklist
+
+- Treat this as a AI Product And UX task centered on annotation workflows.
+- Start from user job, risk, and feedback loop.
+- Expose uncertainty and recovery paths without clutter.
+- Measure usefulness, trust, and operational burden.
+- Prefer existing project conventions, libraries, schemas, and deployment patterns.
+- Avoid broad rewrites unless the current structure blocks a correct solution.
+
+## Deliverables
+
+- Produce implementation steps, code edits, tests, prompts, model choices, data checks, or review findings as the task requires.
+- Make tradeoffs explicit: accuracy, latency, cost, safety, maintainability, and user experience.
+- Leave enough context for another engineer or agent to reproduce the result.

skills/annotation-workflows/agents/openai.yaml

@@ -0,0 +1,3 @@
+display_name: "Annotation Workflows"
+short_description: "Work on annotation workflows for AI Product And UX."
+default_prompt: "Use this skill to help with annotation workflows in AI Product And UX."

skills/anomaly-detection/SKILL.md

@@ -0,0 +1,29 @@
+---
+name: anomaly-detection
+description: "Guidance for anomaly detection in Machine Learning. Use when Codex needs to plan, build, review, test, debug, or document work involving anomaly detection, machine learning, AI systems, software delivery, data workflows, or model quality."
+---
+
+# Anomaly Detection
+
+## Core Workflow
+
+1. Confirm the user goal, target environment, inputs, outputs, constraints, and success criteria for anomaly detection.
+2. Inspect existing code, data, prompts, models, configs, or product flows before proposing changes.
+3. Choose the smallest implementation path that fits the current system and keeps future maintenance clear.
+4. Add focused validation for behavior, quality, security, privacy, cost, and operational risk where relevant.
+5. Report the concrete change, verification performed, and remaining assumptions.
+
+## Subject Checklist
+
+- Treat this as a Machine Learning task centered on anomaly detection.
+- Start with a simple baseline and leakage checks.
+- Keep train, validation, and test boundaries explicit.
+- Report metrics that match the product decision.
+- Prefer existing project conventions, libraries, schemas, and deployment patterns.
+- Avoid broad rewrites unless the current structure blocks a correct solution.
+
+## Deliverables
+
+- Produce implementation steps, code edits, tests, prompts, model choices, data checks, or review findings as the task requires.
+- Make tradeoffs explicit: accuracy, latency, cost, safety, maintainability, and user experience.
+- Leave enough context for another engineer or agent to reproduce the result.

skills/anomaly-detection/agents/openai.yaml

@@ -0,0 +1,3 @@
+display_name: "Anomaly Detection"
+short_description: "Work on anomaly detection for Machine Learning."
+default_prompt: "Use this skill to help with anomaly detection in Machine Learning."

skills/answer-citation-checks/SKILL.md

@@ -0,0 +1,29 @@
+---
+name: answer-citation-checks
+description: "Guidance for answer citation checks in Prompting And Evaluation. Use when Codex needs to plan, build, review, test, debug, or document work involving answer citation checks, prompting and evaluation, AI systems, software delivery, data workflows, or model quality."
+---
+
+# Answer Citation Checks
+
+## Core Workflow
+
+1. Confirm the user goal, target environment, inputs, outputs, constraints, and success criteria for answer citation checks.
+2. Inspect existing code, data, prompts, models, configs, or product flows before proposing changes.
+3. Choose the smallest implementation path that fits the current system and keeps future maintenance clear.
+4. Add focused validation for behavior, quality, security, privacy, cost, and operational risk where relevant.
+5. Report the concrete change, verification performed, and remaining assumptions.
+
+## Subject Checklist
+
+- Treat this as a Prompting And Evaluation task centered on answer citation checks.
+- Create representative positive, negative, and adversarial examples.
+- Separate prompt changes from model changes during evaluation.
+- Use measurable rubrics instead of vague quality language.
+- Prefer existing project conventions, libraries, schemas, and deployment patterns.
+- Avoid broad rewrites unless the current structure blocks a correct solution.
+
+## Deliverables
+
+- Produce implementation steps, code edits, tests, prompts, model choices, data checks, or review findings as the task requires.
+- Make tradeoffs explicit: accuracy, latency, cost, safety, maintainability, and user experience.
+- Leave enough context for another engineer or agent to reproduce the result.