abgregs/skills

Letzte erfasste Quellaktivität: 28. Oktober 2026 um 03:55

SkillsMP-Katalog aktualisiert: 5. November 2026 um 19:57

gesammelte Skills: 6

GitHub-Stars: 0
GitHub-Forks: 0

Mit einem KI-Assistenten installieren

Diesen Prompt kopieren und in den verwendeten KI-Assistenten einfügen.

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

README des Repositorys

README-Quelldatei: README.md

README-Quelldatei lesen

Vollständige README anzeigen

# skills A collection of AI coding skills organized by skill group. Each group is a cohesive set of skills designed to work together. ## Skill Groups | Group | Skills | Description | |-------|--------|-------------| | [doc-driven-dev](./doc-driven-dev/) | `/brief`, `/debrief`, `/setup-docs` | Keep docs aligned with code | | [git](./git/) | `/git-commit`, `/git-create-pr` | Commits that each stand alone, and pull requests written from the diff | | [scope-shaping](./scope-shaping/) | `/sweep` | Widen narrowly-stated tasks to cover related work in the same domain | ## Install ### Via skills CLI (any supported agent) Install all skills (interactive picker): ```bash npx skills@latest add abgregs/skills ``` Install only a specific group: ```bash npx skills@latest add abgregs/skills/doc-driven-dev ``` Install a single skill by name: ```bash npx skills@latest add abgregs/skills@brief ``` ### Manual (Claude Code only) Clone and link all skill groups: ```bash git clone https://github.com/abgregs/skills.git cd skills ./scripts/link-skills.sh ``` Or link a specific group: ```bash ./scripts/link-skills.sh doc-driven-dev ``` ## Structure ``` skills/ ├── README.md # this file ├── .claude-plugin/ │ └── plugin.json # skill registry (all groups) ├── scripts/ │ └── link-skills.sh # per-group or all-group linking │ ├── doc-driven-dev/ # skill group │ ├── README.md │ └── skills/ │ ├── brief/ │ ├── debrief/ │ └── setup-docs/ │ ├── git/ # skill group │ ├── README.md │ └── skills/ │ ├── git-commit/ │ └── git-create-pr/ │ └── scope-shaping/ # skill group ├── README.md └── skills/ └── sweep/ ``` ## Skill layout Every skill folder follows one layout. `SKILL.md` is what the agent loads; everything else keeps it short and gives humans a way in. ``` <skill-name>/ ├── SKILL.md # required — frontmatter + the instructions the agent runs ├── README.md # required — what it does, how to invoke it, what it produces ├── LICENSE.md # required — MIT, Austin Gregersen ├── NOTICE.md # optional — legal notices: third-party IP, trademarks, attributions ├── scripts/ # deterministic steps SKILL.md invokes (detector + judge, see CLAUDE.md) ├── examples/ # worked examples: sample invocations, sample output, filled templates └── references/ # background the agent consults on demand: research, data, specs, rationale ``` **The pattern matters more than the two named folders.** `SKILL.md` carries only what the agent needs on every run; anything it reaches for sometimes — a worked example, a lookup table, research notes, collected data — moves into its own folder at the skill root and `SKILL.md` points at it. `examples/` and `references/` are the two cases that come up most often; add another folder when a skill accumulates a different kind of material. - `examples/` and `references/` are optional but encouraged: they carry context that would otherwise bloat `SKILL.md` or the README, and they keep a skill's material organized where the next edit can find it. - `README.md` is for people: what the skill does, how to invoke it, what it produces, and what else is in the folder. `SKILL.md` is for the agent. Do not restate one in the other. - `NOTICE.md` appears only when there is something to notice: a skill that embeds or depends on someone else's intellectual property, a registered trademark, or material with its own attribution terms. - `LICENSE.md` is the same MIT text in every skill, so each one stays usable when installed on its own. - Scripts are referenced from `SKILL.md` via the skill's base directory (`<skill-base-dir>/scripts/...`), never a hardcoded install path. ## Adding a New Skill Group 1. Create a directory at the repo root with a `README.md` and `skills/` subdirectory 2. Add each skill as a folder under `skills/` using the layout above — at minimum `SKILL.md` (with `name` and `description` frontmatter), `README.md`, and `LICENSE.md` 3. Register each skill's path in `.claude-plugin/plugin.json` 4. Add the group to the table above ## Key Considerations - **Naming collisions**: Skill names must be unique across all groups. Both the skills CLI and the link script install into flat directories, so two skills with the same name would collide. The link script warns if a collision is detected. - **Shared utilities**: If groups ever need shared helper logic, add a top-level `shared/` directory. Defer this until actually needed. ## License MIT — see [LICENSE.md](./LICENSE.md). Each skill also carries its own copy.

README des Repositorys wird geladen…

Beiträge

Alle ansehen →

abgregs

@abgregs5. Nov. 2026 05:51 UTC

Beitrag zu einem Skill-ReleaseGestapelte PRs verwalten

Beitragszusammenfassung · Englisch

Groups diffs into lint-gated commits in plain technical English, using scripts to preflight, validate messages, and confirm each commit.

Originalbeitrag

Skill-Details ansehen · skills/git-skills-git-commit

Skills in diesem Repository

Klassifizierung ausstehend

Skill

Beruf

Beschreibung

Aktualisiert

git-commit

nicht klassifiziert

Generate conventional commit messages and execute commits autonomously. Use when the user is committing code or asks for a commit message. Commits staged changes as one commit (single mode), splits unstaged changes into logical commits (grouped mode), or…

Quellsprache: Englisch

2026-10-28

git-create-pr

nicht klassifiziert

Push the current branch and open a pull request through gh, with the body generated from the diff. Use when the user asks to open, create, or update a PR, or to push a branch for review.

Quellsprache: Englisch

2026-10-13

brief

nicht klassifiziert

Pre-task briefing. Searches project docs for relevant conventions, rules, and requirements before planning code changes. Use when starting any non-trivial task involving code changes.

Quellsprache: Englisch

2026-10-07

debrief

nicht klassifiziert

Post-task debrief. Reviews code changes against project docs, proposes doc updates for new patterns or convention changes, and maintains doc health. Use after completing any non-trivial code task.

Quellsprache: Englisch

2026-10-07

setup-docs

nicht klassifiziert

Bootstrap or audit the docs/ folder structure for a project. Explores the codebase, interviews the user about conventions, and builds out documentation incrementally. Use when starting a new project or when docs need a major refresh.

Quellsprache: Englisch

2026-10-07

sweep

nicht klassifiziert

Scope-expansion skill. Identifies near-identical sibling tasks in the same domain as the stated task, audits existing patterns for consistency, and produces an expanded plan covering original + obvious + borderline siblings. Use when a stated task lives in a…

Quellsprache: Englisch

2026-10-07

Es werden 6 von 6 gesammelten Skills angezeigt.