refined-agent-skills
A curated collection of high-signal engineering and professional agent skills distilled from real collaborative sessions.
Open source Open in the app JSON README (API)
About
A curated collection of high-signal engineering and professional agent skills distilled from real collaborative sessions.
Details
- Kind
- Plugins
- Topic
- No topic detected
- Publisher
- jsebayhi
- Origin
- gemini
- Category
- ferramentas
- Version
- 0.3.0
- Open pull requests
- 1
- Last push
- 2026-03-25T13:19:43Z
- Repository state
- ativo
- Language
- JavaScript
- License
- MIT
- Added
- 2026-08-30 14:13:39
- Updated
- 2026-08-30 14:13:39
- Origin id
jsebayhi/refined-agent-skills
README
# Refined Agent Skills (refined-agent-skills)
A collection of high-signal engineering and professional agent skills for the Gemini CLI. This extension provides a curated suite of "distilled experts" designed to build grounded, reliable, and maintainable software through senior-level protocols.
## Installation
To make all the skills in this repository available in your Gemini CLI at once:
```bash
# Install from a remote git repository
gemini extensions install <git-repo-url>
# Or link the entire repository of skills locally (updates reflect immediately)
gemini skills link /path/to/refined-agent-skills
```
---
## 🏗️ The Production-Grade Software Lifecycle
This suite follows an **"Orchestrator + Specialized Experts"** architecture. For complex tasks, agents are encouraged to activate multiple skills simultaneously.
### 1. Orchestration (The Framework)
* **[Orchestrating Software Lifecycle](skills/orchestrating-software-lifecycle/SKILL.md)**: Manages the full 5-phase SDLC (Alignment, Architecture, Implementation, Validation, Submission).
* *Triggers*: "fix a bug", "implement a feature", "address PR feedback".
### 2. Strategy & Architecture (The "Why")
* **[Deep Brainstorming](skills/deep-brainstorming/SKILL.md)**: Divergent thinking using Tree of Thought and Red Teaming during Phase 2.
* *Triggers*: "brainstorm", "explore possibilities", "think deeply".
* **[Maintaining Rigorous Architecture Decisions](skills/maintaining-rigorous-architecture-decisions/SKILL.md)**: Convergent codification of decisions using ADRs. Mandates 3+ alternatives.
* *Triggers*: "create an ADR", "architectural decision", "analyze alternatives".
### 3. Implementation Experts (The "Hands")
* **[Engineering Reliable Software with Python](skills/engineering-reliable-software-with-python/SKILL.md)**: Production-grade Python, Ruff/Black compliance, and Pytest implementation.
* *Triggers*: "write a python script", "fix the python code", "test with pytest".
* **[Engineering Reliable Scripts with Bash](skills/engineering-reliable-scripts-with-bash/SKILL.md)**: Robust shell scripting, ShellCheck compliance, and BATS testing.
* *Triggers*: "write a bash script", "fix the shell script", "test with BATS".
* **[Authoring High-Signal Git Commits](skills/authoring-high-signal-git-commits/SKILL.md)**: Generates descriptive, truth-based commit messages from staged changes.
* *Triggers*: "git commit", "prepare commit", "commit changes conventionally".
### 4. Quality & Testing (The "Validation")
* **[Testing Software Efficiently](skills/testing-software-efficiently/SKILL.md)**: The language-agnostic "Testing Trophy" methodology.
* *Triggers*: "testing strategy", "testing trophy", "mocking philosophy".
* **[Validating User Interfaces](skills/validating-user-interfaces/SKILL.md)**: Enforces POM, signal-based waiting, and user-centric assertions for UI/E2E testing.
* *Triggers*: "write a ui test", "e2e testing", "playwright tests".
### 5. Infrastructure & Security (The "Hardening")
* **[Upholding DevSecOps Standards](skills/upholding-devsecops-standards/SKILL.md)**: Shift-Left security, secret detection, and Local-CI Alignment.
* *Triggers*: "security audit", "harden the pipeline", "check for secrets", "local ci alignment".
### 6. Delivery & Collaboration (The "Submission")
* **[Collaborating on Git Projects](skills/collaborating-on-git-projects/SKILL.md)**: Dual-platform (gh/glab) support, Conventional Commits, and PRB rationale.
* *Triggers*: "create a PR", "open a merge request", "PR standards".
* **[Interacting with GitHub](skills/interacting-with-github/SKILL.md)**: Production-grade GitHub management using `github:*` MCP tools. Supports precise PR feedback, security auditing, and GitHub Actions monitoring.
* *Triggers*: "create a PR", "start a GitHub review", "check the actions", "search GitHub".
* **[Interacting with GitLab](skills/interacting-with-gitlab/SKILL.md)**: Production-grade GitLab management using `gitlab:*` MCP tools. Supports precise line-level feedback, security auditing, and pipeline monitoring.
* *Triggers*: "comment on a line", "start a GitLab review", "check vulnerabilities", "answer a discussion", "check the pipeline".
* **[Authoring Effective User Documentation](skills/authoring-effective-user-documentation/SKILL.md)**: Enforces Diátaxis and "Documentation is Code."
* *Triggers*: "update the readme", "write documentation", "create a user guide".
### 7. External Research & Discovery
* **[Investigating External Dependencies](skills/investigating-external-dependencies/SKILL.md)**: Senior-Level library investigation (source code, APIs, best practices).
* *Triggers*: "how does this library handle [X]?", "weird error from [Dependency]", "discover the grain".
### 8. Agentic Systems & Architecture
* **[Engineering Multi-Agent Systems](skills/engineering-multi-agent-systems/SKILL.md)**: The "Standard Model" of agentic architecture, topologies, and cognitive thermodynamics.
* *Triggers*: "multi-agent system", "synergistic agents", "agentic architecture", "coordination tax".
---
## 🛠️ Meta-Engineering
* **[Engineering Agent Skills](skills/engineering-agent-skills/SKILL.md)**: The expert system for designing and scaffolding new agent skills.
* *Triggers*: "create a skill", "build a workflow", "design an agent tool".
---
## 🔍 Analysis & Research Utilities
* **[Performing Systematic Root Cause Analysis](skills/performing-systematic-root-cause-analysis/SKILL.md)**: The "Debug Detective" hypothesis-driven workflow. Enforces a strict 5-attempt threshold before mandatory checkpoint backtracking.
* *Triggers*: "investigate", "bug", "error", "failure", "root cause".
* **[Deep Reading Agent Skill](skills/deep-reading-agent-skill/SKILL.md)**: Section-by-section analysis and reflection on complex resources.
* *Triggers*: "read deeply", "analyze the nuances", "reflect on the spirit".