igorrendulic/engineering-skills

Última atividade de origem registrada: 15 de outubro de 2026 às 20:00

Catálogo do SkillsMP atualizado: 2 de novembro de 2026 às 18:02

skills coletadas: 8

Estrelas no GitHub: 1
Forks no GitHub: 0

Instalar com um assistente de IA

Copie este prompt para o assistente de IA que você usa.

Follow https://skillsmp.com/skill-install/prompt.md to install Agent Skills from https://github.com/igorrendulic/engineering-skills.

README do repositório

Arquivo de origem do README: README.md

Ler o README original

Expandir README completo

# Engineering Skills A small set of AI coding-agent skills for taking work from an idea to a verified pull request. Plan around existing code, implement in manageable increments, and review both correctness and structure. **Simple designs. Explicit scope. Evidence before completion.** ## Quick start Install in Codex: ```sh codex plugin marketplace add igorrendulic/engineering-skills --ref master && \ codex plugin add engineering-skills@engineering-skills ``` Start a new Codex session after installing. Invoke a workflow or ask for help: ```text $engineering-skills:help $engineering-skills:help plan $engineering-skills:init Initialize this project's AGENTS.md $engineering-skills:plan Add password reset $engineering-skills:implementation Implement the password-reset plan $engineering-skills:pr-review Review the password-reset changes $engineering-skills:github-pr Create a PR for these changes ``` Help explains workflows, inputs, outputs, and examples without executing them. The workflow examples are agent prompts, not shell commands. The package preserves shared references and sibling skills. `plan` includes structural design review. `implementation` includes structural review at substantial checkpoints and independent `pr-review` when required, with repairs and re-review. The example makes `pr-review` explicit before PR creation; an existing review can be reused while its scope, snapshot, and evidence still apply. ## The SDLC ```mermaid flowchart TD A["init: establish project instructions"] --> B subgraph Planning["plan workflow"] B["Define goal, scope, and design"] --> C["structural-review: assess the design"] C --> D["Resolve findings; define tasks and verification"] D --> P{"Plan ready?"} P -->|Revise within scope| B P -->|Blocked| Q["Report blockers; plan remains not ready"] end P -->|Yes, implementation authorized| E subgraph Implementation["implementation workflow"] E["Build or repair and verify increments"] --> F["structural-review: at substantial checkpoints"] F --> S{"Required structural review complete, with no unresolved High or Medium findings?"} S -->|Yes or not required| G["Verify the complete outcome"] S -->|No| R G --> H{"Independent review required?"} H -->|Yes| I["pr-review: fresh-context correctness review"] I --> J{"No blocking findings and required verification satisfied?"} H -->|No| J J -->|No| R{"Can resolve within scope and remaining repair limit?"} R -->|Yes| E R -->|No: limit, no progress, or blocked| X["Preserve work; report unresolved findings or evidence gaps"] J -->|Yes| V["Retain verified implementation; report results"] end V -->|PR delivery authorized| K["github-pr: create or update the PR"] ``` Start at the stage that fits the work; small, obvious changes can go straight to implementation. Review steps inside each workflow are invoked by that workflow. Independent review is required for substantial behavior changes, changes affecting security, data integrity, or shared/public contracts, and whenever the user or repository requires it. Explicit review waivers must be disclosed. The review-and-fix loop defaults to at most three repair rounds, each including a fresh follow-up review; it stops earlier when progress or authorization is blocked. Missing required evidence prevents completion. Publishing, merging, and deployment follow the user's authorization. ## Workflows | Workflow | What it does | | --- | --- | | [init](skills/init/SKILL.md) | Create or conservatively update root `AGENTS.md` from repository evidence. | | [plan](skills/plan/SKILL.md) | Define success criteria, review the design, and order verifiable tasks. | | [implementation](skills/implementation/SKILL.md) | Implement, verify, and repair supported findings within scope. | | [structural-review](skills/structural-review/SKILL.md) | Assess ownership, complexity, reuse, and maintainability in plans or code. | | [pr-review](skills/pr-review/SKILL.md) | Independently review correctness and regressions in fresh context. | | [github-pr](skills/github-pr/SKILL.md) | Prepare branches, commits, and PRs with clear rationale and validation evidence. | | [help](skills/help/SKILL.md) | Explain usage and workflow selection without executing a workflow. | Plans default to `tasks/<feature-slug>/plan.md` and `todo.md`, while respecting existing project conventions. See the shared [plan lifecycle](references/plan-lifecycle.md) and [code quality principles](references/code-quality.md). ## Contributing Add workflows under `skills/<name>/SKILL.md` and register them in the root [dispatcher](SKILL.md). Keep instructions focused, reuse shared references, and include verification appropriate to the change. The dispatcher table is also the help catalog. Keep skill names short; the plugin manifest owns the namespace. See [local checks](evals/README.md) for package loading, installer regression tests, and skill evaluation. Bump the manifest version when releasing changed plugin contents so installed versions remain identifiable.

Carregando o README do repositório…

Skills neste repositório

classificação pendente

skill

ocupação

descrição

atualizado

engineering-skills

sem classificação

Route engineering-skills requests to bundled initialization, planning, implementation, review, PR, or usage-help workflows. Use when the user invokes the standalone engineering-skills dispatcher or asks which bundled workflow to use.