Update (2026-07-21): Synchronized the current names and ownership boundaries for all 15 Skills, including
product-spec,ui-design, the implementation Skills, and the authorization boundaries for review, delivery, and external research.
idaibin/skills is a collection of independent Agent Skills for real software engineering tasks. It no longer acts as a prompt repository, shared-rules center, or mandatory workflow framework.
Each Skill owns its trigger, workflow, authority boundary, output contract, references, and evaluation cases. A Skill can be installed on its own. Several Skills may be composed when a task crosses owners, but they do not form a required end-to-end state machine.
Current catalog
Repository engineering
| Skill | Use when |
|---|---|
repo-map |
Repository boundaries, commands, task routes, or reusable owners need a verified map |
domain-modeling |
Shared cross-functional business terms or rules conflict; lifecycle and bounded contexts are conditional depth |
repo-review |
The current Worktree/index, a fixed SHA/range, or a verified review package needs read-only review |
repo-delivery |
Reviewed changes need categorized commits by default, an explicit single commit, push/sync, evidence-based branch integration, or cleanup |
Product and design
| Skill | Use when |
|---|---|
product-spec |
User-visible behavior, rules, states, and acceptance need an implementation-ready product specification without API design |
ui-design |
One page or flow needs Feature UI by default; shared tokens, component semantics, or visual language activate the conditional Design System profile |
Implementation
| Skill | Use when |
|---|---|
dev-frontend |
A frontend change must be implemented and verified within the existing stack and UI constraints |
dev-rust |
A Rust change or refactor must be implemented against the real toolchain and applicable risk boundaries |
Specialist audit and runtime verification
| Skill | Use when |
|---|---|
audit-frontend |
A known frontend surface needs architecture, accessibility, state, performance, or design review |
audit-rust |
Rust ownership, concurrency, SQLite, unsafe/FFI, performance, or memory needs bounded review |
audit-security |
A known security-sensitive code or configuration surface needs a read-only assessment |
ops-browser |
A browser target must be operated after target, identity, and authorization checks with before/after evidence |
ops-client |
A Tauri, Electron, or native desktop client must be verified against its real process and window |
Extensions
| Skill | Use when |
|---|---|
ask-chatgpt |
Codex needs to ask ChatGPT for an explicitly authorized independent review, research result, visual exploration, or decision challenge |
human-writing |
Source-grounded prose must be drafted or revised while preserving facts, attribution, voice, and disclosures |
Planning and Diagnosis are no longer standalone Skills. Complex work uses the host agent’s planning capability and effective repository rules. Concrete failures follow the evidence-driven diagnosis requirements in personal or repository instructions. A capability remains a Skill only when it needs its own references, workflow, or specialized validation.
Install
Browse and select interactively:
npx skills@latest add idaibin/skills
List the current catalog:
npx skills@latest add idaibin/skills --list
Install selected Skills globally for Codex:
npx skills@latest add idaibin/skills \
--skill repo-map domain-modeling repo-review \
--global --agent codex
Treat the GitHub repository and skills.sh page as the current catalog and installation sources. The blog explains the catalog; it does not mirror every Skill package in full.