lefant/agent-skills

最近记录的来源活动: 2026年10月30日 19:05

SkillsMP 收录数据更新: 2026年11月12日 12:52

已收集 skills: 70

GitHub 星标: 1
GitHub Forks: 1

用 AI 助手安装

复制这段 Prompt，发给你正在使用的 AI 助手。

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

仓库 README

仓库 README 来源：README.md

查看 README 源文件

展开完整 README

# lefant/agent-skills Curated skills for AI coding agents (Claude Code, Codex, OpenCode). ## Structure - `lefant/` - Custom skills developed in-house - `vendor/` - Vendored skills from upstream repositories (reviewed for security) ## Reference Guides - [Adopt Documentation & Process](docs/reference/adopt-documentation-process.md) — A single-file instruction for adding a documentation map and development workflow to another repository's README, with pointers from AGENTS.md and CLAUDE.md. ## Documentation & Process ### Documentation (`./docs`) Use these documentation areas. Create optional directories only when there is content to own; not every area below exists in this tooling repository yet. - **`docs/specs/`**: Feature Specifications — living documents describing WHAT features do and WHY - Skill: [`feature-specs`](lefant/feature-specs/SKILL.md) - **`docs/decisions/`**: Architecture Decision Records (ADRs) — immutable records of architectural choices once accepted; supersede rather than rewrite their rationale - Skill: [`architecture-decision-records`](lefant/architecture-decision-records/SKILL.md) - **`docs/changelog/`**: Release notes and changelog fragments - Skill: [`changelog-fragments`](lefant/changelog-fragments/SKILL.md) - **`docs/solutions/`**: Documented solutions to past problems and team learnings, organized by category with YAML frontmatter (`module`, `tags`, `problem_type`) - **`docs/reference/`**: Technical references and operational details - **`docs/api/`**: Technical references for internal APIs, when applicable - **`docs/brand-guide.md`**: Brand and design guide, when applicable - **`docs/research/`**: Discovery notes and analysis - **`docs/brainstorms/`**: Existing standalone brainstorms and design explorations - **`docs/plans/`**: Time-stamped implementation plans - **`docs/devlog/`**: Time-stamped implementation outcomes, caveats, and learnings - Skill: [`devlog`](lefant/devlog/SKILL.md) The spec, ADR, and changelog skills replace the old `spec_create`, `adr_create`, and `changelog_create` commands. Use the installed agent's skill invocation syntax; the links above point to the maintained skill instructions. ### AI-Assisted Development (`./docs`) **Workflow**: Research/Plan → Implement → Review → Compound Prefer the Compound Engineering skills when installed. Verify the installed version's skill names and output paths before invoking them; this repository does not install that plugin. The workflow also works without it. Commit progress continuously in small, reviewable increments. Once continuous pushes to the agreed remote and branch are authorized, push each coherent, checked checkpoint without asking again. Inspect automatic deployment effects first; separate release and shared-state actions still need their own authorization. 1. **Research/Plan** — document current behavior, constraints, and requirements; define implementation strategy and verification. Durable research belongs in `docs/research/`, and plans in `docs/plans/`. A plugin may keep brainstorm and planning work in one plan artifact; do not duplicate it just to fill folders. 2. **Implement** — make the source changes and verify them against the plan. 3. **Review** — inspect correctness, contracts, security, and test coverage; resolve findings and rerun affected checks. 4. **Compound** — capture verified reusable solutions in `docs/solutions/` and update their existing owners rather than duplicating knowledge. At meaningful checkpoints, use [`devlog`](lefant/devlog/SKILL.md) to record outcomes, evidence, caveats, and follow-ups in `docs/devlog/`. Devlogs preserve session history; solutions preserve reusable knowledge. Use [`atomically-land`](lefant/atomically-land/SKILL.md) when closing out work that also needs plans, specs, ADRs, or changelog updates. Use `docs/research/`, `docs/plans/`, and `docs/devlog/` for new work. The [`rpi`](lefant/rpi/SKILL.md) skill uses these documentation destinations while preserving its explicitly requested legacy workflows; it is not the default process. Historical documents may retain former `thoughts/shared/` paths as provenance. Do not move historical files merely to follow the new convention. ## Usage ### With skills CLI ```bash skills add /path/to/agent-skills --skill '*' -g -y -a claude-code -a codex -a opencode ``` Wildcard installs include optional packages. Caveman, ZFC, and Pulumi are [explicit opt-in / repository-optional](vendor/README.md#opt-in--repository-optional-skills), not global defaults; select individual skills instead of `'*'` to omit them. ### In Docker Cloned to `/opt/lefant-agent-skills` and symlinked via entrypoint. ## Updating Vendored Skills ```bash ./scripts/update-vendor.sh ``` Then review changes before committing. For a targeted refresh, pass upstream repository names, for example: ```bash ./scripts/update-vendor.sh antithesishq/antithesis-skills typesafe-ai/skills ``` The updater stages fetches, local patches, and checks before replacing `vendor/`. Failures leave the existing snapshot unchanged. See the [vendor maintenance notes](vendor/README.md#update-process) for held updates. ### Vendor Quality Checks Before committing vendor changes, verify skill discoverability: ```bash ./scripts/check-vendor-layout.sh python3 scripts/tests/vendor-safety.test.py -v ``` For quick manual inspection: ```bash find vendor -type d -name skills -print find vendor -path '*/SKILL.md' -printf '%h\n' | sort ``` Rules: - `find vendor -type d -name skills -print` should normally print nothing. Vendored skills must be flattened to `vendor/<source>/<skill>/SKILL.md`, even when upstream stores them under `skills/`. - Do not vendor whole plugin bundles that need agents, commands, MCP config, or other non-skill assets unless the installer can consume those assets. Handle plugin-native bundles in a plugin/package layer instead. - Keep source-to-local path mappings in `scripts/update-vendor.sh`; do not rely on symlinks for discoverability. - Keep frontmatter `name:` values unique across `lefant/` and `vendor/`; installers expose one skill per name. ## Skills Included ### Custom Skills (`lefant/`) | Skill | Description | |-------|-------------| | `github-access` | Access GitHub repositories via gh CLI or REST API | | `sentry` | Fetch and analyze Sentry issues, events, and logs | | `architecture-decision-records` | Create and manage ADRs | | `changelog-fragments` | Maintain changelog fragments for conflict-free history | | `feature-specs` | Document feature requirements and scenarios | | `atomically-land` | Close out implementation work and update surrounding docs/tracking | | `devlog` | Write implementation devlogs for current session state | | `exa` | Use Exa for web search, code context lookup, and current-information lookups via SDK, API, or MCP | | `skills-best-practices` | Create, review, and refactor agent skills with self-contained packaging and eval guidance | | `github-get-pr-comments` | Gather PR comments and combine them with local project context | | `git-resolve-merge-conflicts` | Resolve git merge conflicts safely using local context | | `mermaid-diagrams` | Create hierarchical Mermaid diagrams | | `handover` | Generate a concise resume prompt for the next session | | `hitl-browser-automation` | Run human-in-the-loop browser automation with VNC, loopback CDP, smoke validation, and agent-browser continuation | | `rpi` | Preserve the create_plan, implement_plan, and research_codebase workflows | | `qrspi` | Portable adaptation of matanshavit/qrspi's eight-phase workflow; upstream provenance and local changes documented in `lefant/qrspi/SOURCE.md` | | `recent-context-from-git` | Summarize recent local docs and work context from git history | | `test-analyzer` | Analyze CTRF test reports with jq | | `expect-testing` | Create and review expect tests, inline snapshots, approval tests, and golden files | | `youtube-transcript` | Fetch YouTube video transcripts | | `tasknotes` | Manage Obsidian tasks through maintained TaskNotes interfaces or a safe Markdown fallback | | `untis-access` | Access WebUntis data, compare repo implementations, and run proven message/timetable scripts | ### Vendored Skills (`vendor/`) See `vendor/README.md` for sources and update process.

正在加载仓库 README…

这个仓库中的 skills

软件开发工程师网络与计算机系统管理员平面设计师技术写作员

7 个职业分类 · 已分类 36%

skill

职业分类

描述

更新

hitl-browser-automation

软件开发工程师

Use when browser automation needs a human checkpoint: the user wants to demonstrate a workflow once, watch and verify an agent-driven browser result before proceeding, log in or complete MFA/credential-sensitive steps manually, connect through VNC, or use…

原文语言：英语

2026-10-30

marimo-pair

未分类

Drive a live marimo notebook as a workspace: run Python in the same kernel the user does, inspect live notebook state, and commit durable notebook changes. Use when the user wants to start a marimo notebook or pair on an active marimo session.

原文语言：英语

2026-10-30

antithesis-debug

未分类

Run a bash script inside an Antithesis run at a chosen moment with `snouty runs exec`, from the terminal and without a browser. Use it for scripted forensics on a run: process and container inspection, on-disk evidence, gdb attach, and general debugging.…

原文语言：英语

2026-10-30

antithesis-documentation

未分类

Use Antithesis documentation efficiently for product, workflow, and integration questions. Prefer the snouty docs CLI when available, and otherwise request markdown versions of documentation pages directly.

原文语言：英语

2026-10-30

antithesis-feature-workload

未分类

Design and build an Antithesis workload for a specific feature: analyze the feature, discover its testable properties, build a self-driving workload that runs locally and in Antithesis, and iterate until the feature's behaviors are exercised and its…

原文语言：英语

2026-10-30

antithesis-launch

未分类

Launch an Antithesis run with snouty by discovering the harness layout, building the right Docker Compose config, running `snouty validate`, bailing on validation failure, and then submitting `snouty launch` with sane metadata. Use when the user wants to…

原文语言：英语

2026-10-30

antithesis-mutation-testing

未分类

Validate that an Antithesis property catalog can actually catch bugs. For each safety property, inject one realistic mutant into the SUT, run it, and confirm the targeted property fires. Diagnoses survivors as a bad mutant, bad oracle, bad workload, or bad…

原文语言：英语

2026-10-30

antithesis-research

未分类

Analyze a codebase to figure out how it should be tested with Antithesis: map the system, identify failure-prone areas and testable properties, and produce the research artifacts needed for workload and environment planning.

原文语言：英语

2026-10-30

antithesis-review-inputs

未分类

Review a workload's input generation for state space exploration: how well it uses randomness to drive the system under test into diverse regions of behavior.

原文语言：英语

2026-10-30

antithesis-setup-k8s

未分类

Scaffold the Antithesis harness for customers running Kubernetes: Convert helm charts, kustomize templates, or raw Kubernetes manifests into a minimized set of manifests for Antithesis. Verify the Kubernetes manifests are in the correct shape with snouty…

原文语言：英语

2026-10-30

antithesis-setup

未分类

Scaffold the Antithesis harness with docker-compose: initialize the working directory, write Dockerfiles and docker-compose.yaml with build directives, and prepare to submit your first Antithesis test run. If the desired setup is Kubernetes, defer to the…

原文语言：英语

2026-10-30

antithesis-skills-feedback

未分类

File a bug report or feedback on the Antithesis skills by opening a pre-filled GitHub issue URL (skill name, skill version, agent, short summary). Load when the user wants to report a problem with, or give feedback on, any Antithesis skill. Does not…

原文语言：英语

2026-10-30

antithesis-triage

未分类

Triage Antithesis test reports to understand what happened in a run: look up runs, check status, investigate failed properties (assertions), view metadata, download logs, search events across all timelines, inspect findings, and examine environmental details.…

原文语言：英语

2026-10-30

antithesis-workload

未分类

Implement Antithesis workloads by turning the property catalog into SDK assertions and test commands, then refine coverage after triage.

原文语言：英语

2026-10-30

hegel

未分类

Write property-based tests using Hegel across Rust, Go, C++, TypeScript, Java, OCaml, and Python projects. Use this skill whenever the user asks to write tests, add test coverage, or improve testing for functions, modules, or libraries — especially when the…

原文语言：英语

2026-10-30

caveman-commit

未分类

Terse Conventional Commit messages; does not stage or commit. Use only when explicitly requested by name or enabled in repository instructions.

原文语言：英语

2026-10-30

caveman-compress

未分类

Compress prose memory files into Caveman style with backups. Use only when explicitly requested by name or enabled in repository instructions.

原文语言：英语

2026-10-30

caveman-help

未分类

Quick reference for Caveman modes and commands; does not activate them. Use only when explicitly requested by name or enabled in repository instructions.

原文语言：英语

2026-10-30

caveman-review

未分类

Terse code review comments; does not post or approve reviews. Use only when explicitly requested by name or enabled in repository instructions.

原文语言：英语

2026-10-30

caveman

未分类

Caveman-style responses, lite by default, with full, ultra, and wenyan modes. Use only when explicitly requested by name or enabled in repository instructions.

原文语言：英语

2026-10-30

zfc

未分类

Zero Framework Cognition principles for projects that explicitly adopt ZFC. Use only when explicitly requested by name or enabled in repository instructions.

原文语言：英语

2026-10-30

librarian

未分类

Cache and refresh remote git repositories under ~/.cache/checkouts/<host>/<org>/<repo> so future references can reuse a local copy. Use this skill when the user points you to a remote git repository as reference or you encountered a remote git repo through…

原文语言：英语

2026-10-30

cloudformation-to-pulumi

未分类

Migrate AWS CloudFormation templates or stacks to Pulumi. Use only when explicitly requested by name or enabled in repository instructions.

原文语言：英语

2026-10-30

package-usage

未分类

Audit package versions across stacks in a Pulumi organization. Use only when explicitly requested by name or enabled in repository instructions.

原文语言：英语

2026-10-30

provider-upgrade

未分类

Upgrade provider SDK dependencies in a Pulumi project and reconcile preview diffs. Use only when explicitly requested by name or enabled in repository instructions.

原文语言：英语

2026-10-30

pulumi-arm-to-pulumi

未分类

Migrate Azure ARM or Bicep templates to Pulumi. Use only when explicitly requested by name or enabled in repository instructions.

原文语言：英语

2026-10-30

pulumi-automation-api

未分类

Embed Pulumi and orchestrate stacks with Automation API. Use only when explicitly requested by name or enabled in repository instructions.

原文语言：英语

2026-10-30

pulumi-best-practices

未分类

Pulumi TypeScript/Python resource, Output, configuration, and CI practices. Use only when explicitly requested by name or enabled in repository instructions.

原文语言：英语

2026-10-30

pulumi-cdk-to-pulumi

未分类

Migrate AWS CDK applications to Pulumi. Use only when explicitly requested by name or enabled in repository instructions.

原文语言：英语

2026-10-30

pulumi-component

未分类

Author and distribute Pulumi ComponentResource classes. Use only when explicitly requested by name or enabled in repository instructions.

原文语言：英语

2026-10-30

pulumi-esc

未分类

Manage environments, secrets, and configuration with Pulumi ESC. Use only when explicitly requested by name or enabled in repository instructions.

原文语言：英语

2026-10-30

pulumi-neo-handoff

未分类

Transfer work to Pulumi Neo only with explicit user handoff authorization. Use only when explicitly requested by name or enabled in repository instructions.

原文语言：英语

2026-10-30

pulumi-overview

未分类

Pulumi CLI, infrastructure-as-code, and operations guidance. Use only when explicitly requested by name or enabled in repository instructions.

原文语言：英语

2026-10-30

pulumi-terraform-to-pulumi

未分类

Translate Terraform/OpenTofu to Pulumi or import Terraform state. Use only when explicitly requested by name or enabled in repository instructions.

原文语言：英语

2026-10-30

pulumi-upgrade-provider

未分类

Maintain Pulumi provider repositories with the upgrade-provider tool. Use only when explicitly requested by name or enabled in repository instructions.

原文语言：英语

2026-10-30

upstream-patches

未分类

Manage Terraform provider submodule patches with scripts/upstream.sh. Use only when explicitly requested by name or enabled in repository instructions.

原文语言：英语

2026-10-30

expect-testing

未分类

Creates and reviews expect tests, inline snapshots, approval tests, and golden files. Use when adding exact-output regression tests or investigating snapshot mismatches. Not for Tcl Expect automation or ordinary expect() assertions.

原文语言：英语

2026-10-30

rpi

未分类

Preserve the RPI workflow commands create_plan, implement_plan, and research_codebase inside a portable skill. Use when the user refers to those exact command names, asks for RPI workflows, or wants the original Claude command behavior preserved as closely as…

原文语言：英语

2026-10-25

changelog-fragments

未分类

Create and manage changelog fragments to track changes without merge conflicts. Each change gets its own file for parallel development.

原文语言：英语

2026-10-24

devlog

未分类

Write implementation devlogs that capture current session progress, learnings, related artifacts, and next steps. Use when the user asks for a devlog or WIP devlog.

原文语言：英语

2026-10-24

已展示 40 / 70 个已收集 Skill。