ilteoood/harness
19 skills19 installs38 starsGitHub
Install
npx skills add https://github.com/ilteoood/harnessSkills in this repo
1CodeqlThe CodeQL skill is a reference for solo builders and small teams managing GitHub code scanning alerts produced by CodeQL analysis. It explains how standard Error, Warning, and Note levels differ from security severities tied to CVSS, when security severity overrides display sorting, and how GitHub derives query severity from CWE-linked CVE score percentiles. Use it while hardening a repo before ship, during PR review when alerts appear, or in operate when backlog triage stalls release cadence. It does not run scans itself; it teaches consistent interpretation so you fix Critical and High items first, handle Notes proportionally, and dismiss with correct rationale. Indie developers shipping SaaS or APIs benefit when they lack a dedicated AppSec reviewer but still need audit-friendly alert handling.1installs2Commit Message StorytellerCommit-message-storyteller is a journey-wide agent skill that helps solo builders write Conventional Commits with bodies that explain motivation, user impact, and risk—not just what files changed. It packages the standard type(scope): description shape plus optional narrative body and footers so changelogs, semantic release tooling, and human reviewers parse history quickly. Examples cover feat, fix, refactor, perf, and docs with realistic scopes like payments, api, and dashboard, showing how to state the bug symptom, the fix, and linked issues. Because every meaningful git commit benefits from the same discipline, the skill applies across Build, Ship, and Operate whenever you stage work, not only at PR time. It is aimed at indie developers who ship alone but still want audit-friendly history and fewer “what does this commit do?” threads. Invoke it after implementing a change set and before git commit or when squashing a branch for merge.1installs3Conventional CommitConventional Commit is an agent skill that walks you through inspecting staged changes and encoding them as Conventional Commits—`type(scope): description` plus optional body and breaking-change footer. Solo and indie builders use it whenever they want git history that release tools, changelogs, and teammates can parse without decoding free-form chat. The skill pairs procedural steps (`git status`, diff review, staging) with an XML-shaped message scaffold so the agent does not invent vague one-liners. It is especially useful when you ship often from a single repo and need consistent scopes across frontend, API, and CI tweaks. It is not a linter or commitizen hook; it is prompt-and-workflow guidance that ends in an explicit `git commit` command pattern. Use it during active feature work and again before opening or updating a pull request.1installs4Create Agentsmdcreate-agentsmd is a procedural skill for solo builders who want coding agents to stop guessing how their repo works. It instructs the agent to author a complete, accurate AGENTS.md at the repository root using the open agents.md format—a dedicated “README for agents” with detailed technical instructions that stay out of human-facing docs. The workflow emphasizes agent-focused content, standard Markdown, flexible sectioning based on project needs, and compatibility across many AI coding tools. Install it when you are bootstrapping a new repo, onboarding agents to an existing codebase, or refreshing context after major stack or workflow changes. The outcome is a single predictable entry point agents can load every session, which reduces wrong assumptions about install steps, test commands, and project conventions compared to ad-hoc prompts or overloading README.md.1installs5Create Implementation PlanCreate-implementation-plan is a Harness skill for solo builders who want specs that agents can execute without clarifying questions. Invoke it when starting a new feature, refactoring, upgrading dependencies, or changing design, architecture, or infrastructure—any time ad-hoc chat planning would leave gaps in file paths and acceptance criteria. The skill mandates fully self-contained plans: phased structure, parallelizable tasks where possible, and zero ambiguous wording. It fits intermediate users who already know what they want built but need the handoff artifact. Primary home is Build PM; it also supports Validate scoping when you need a written execution contract before committing engineering time. Output is meant for autonomous follow-up by other skills or the same agent in execution mode.1installs6Create LlmsCreate LLMs is an agent skill that produces a new llms.txt at the repository root from scratch, following the llms.txt.org specification so large language models have a curated map of what the project is and where to read next. Solo and indie builders shipping open source, internal tools, or agent-facing libraries install it when README alone leaves agents guessing about layout, specs, and high-value paths. The skill mandates reviewing the spec, analyzing the full tree, and discovering README and other anchor files before drafting. That makes it a build-time documentation generator with knock-on benefits at launch for AI-search and agent discoverability. It does not replace a thorough README, API reference generation, or automated sitemap pipelines—you still own accuracy of linked paths and keeping the file updated after refactors.1installs7Create ReadmeCreate-readme is an agent skill that acts as a senior engineer documenting your repository: it reviews the entire workspace, then drafts a comprehensive README.md that is appealing, scannable, and informative. Solo and indie builders use it when a repo is real enough to ship or share but still has a thin or missing README—common right after scaffolding, before a first public push, or when onboarding collaborators. The workflow explicitly benchmarks structure and tone against reference READMEs from Azure Samples and established open-source projects, then keeps prose concise and avoids dumping legal or contribution policy into the README. Formatting targets GitHub Flavored Markdown and community admonition patterns so install steps, architecture notes, and warnings read well on GitHub. You still own accuracy of commands and secrets; the skill focuses on presentation and completeness of the primary entry doc.1installs8Create SpecificationCreate Specification is a harness skill for solo builders and small teams who want one authoritative requirements document before agents write code. It directs you to author a new Markdown spec for a stated purpose, using precise language and structured sections so generative AI can parse requirements, constraints, and interfaces without hunting chat history. Files live in the /spec/ directory with a enforced naming pattern tied to high-level purpose prefixes, which keeps multi-component projects navigable in repos and PRs. The workflow emphasizes self-contained documents with defined terms, tables and lists for structure, and concrete examples including edge cases. It fits early validation when you are scoping a feature or integration, and remains useful during build when you split work across backend, tooling, or infrastructure changes. Treat the output as the contract your agent implements against, then hand off to planning or implementation skills once the spec is stable.1installs9Create Technical SpikeCreate Technical Spike is an agent skill that produces time-boxed technical spike documents for solo and indie builders who need to resolve a specific architecture or integration question before coding. It defines where spikes live on disk, how to name them, and a repeatable markdown skeleton with summary, research questions, deadlines, and ownership metadata so agents do not jump straight into implementation on unresolved bets. Use it when a feature path is blocked on unknowns—API feasibility, realtime performance, third-party coupling—or when you want a written record of what was investigated and what was decided. The skill is procedural knowledge packaged as SKILL.md: your agent fills template inputs (title, category, priority, timebox) and emits decision-ready artifacts for validate and early build planning. It matters because spikes keep small teams from paying full build cost to learn what a week of focused research could have answered.1installs10Create Tldr PageCreate TLDR Page is an agent skill for solo builders who want terminal-friendly cheat sheets without hand-copying from long manuals. It transforms authoritative documentation URLs plus a command identifier into a tldr-pages-compliant page: bounded sections, copy-paste examples, and consistent markdown structure. The workflow enforces prerequisites—if URL or command is missing, it guides you to supply them, optionally fetching the first URL or disambiguating multiple links from a local file. That gatekeeping reduces hallucinated flags and wrong subcommands in shared team repos. Use it when you ship CLIs, internal tools, or homelab utilities and need discoverable references for Cursor, Codex, or Claude Code sessions. It does not replace full tutorials; it compresses what operators reach for daily. Pair with your harness fetch step when docs are only linked inside notes or README fragments.1installs11Dependabotdependabot is a harness reference skill that catalogs every major option for GitHub Dependabot configuration in YAML form. Solo builders maintaining a product repo use it when they need to turn on automated PRs for npm, Cargo, Docker, Composer, or other ecosystems without guessing valid `package-ecosystem` values or schedule syntax. The ingested content is encyclopedic: top-level `version: 2`, optional `registries` for private feeds, `multi-ecosystem-groups` for coordinated update cadence, and per-ecosystem `updates` entries with directories and intervals. It suits indie SaaS and API projects where you wear the DevOps hat and want dependable, reviewable dependency bumps tied to manifest files like `package-lock.json`, `Cargo.lock`, or `Dockerfile`. Pair it with your CI policy so merge rules and test gates match the flood of bot PRs.1installs12Drawiodrawio is a harness-embedded agent skill oriented around turning diagrams.net sources into PNGs that solo builders can drop into repos, PRs, or spec folders. The documented script prioritizes fidelity: use the native draw.io CLI when installed for speed and accuracy, otherwise load the official viewer in a headless browser so orthogonal connectors and container layouts match what authors see in the editor. Batch mode walks directories so a single command refreshes every architecture sheet before a release note or onboarding doc update. For agentic workflows, that means Mermaid-like convenience but with full draw.io expressiveness—provided you accept Node tooling and optional network fetches for the viewer assets. It complements markdown prose rather than replacing diagram authoring; you still maintain .drawio files as the source of truth.1installs13Fastify Best Practicesfastify-best-practices (packaged in harness under an authentication-focused SKILL frontmatter) gives indie API builders concrete Fastify patterns for JWT auth, protected routes, and refresh-token flows. Instead of stitching Stack Overflow snippets, you get TypeScript-oriented register/decorate examples, schema-validated login bodies, and onRequest guards suitable for a first production slice. The skill fits Prism’s Build → backend shelf but honestly supports Ship → security when you harden access control before launch. Use it when you are standing up `/login`, issuing signed claims (id, email, role), and gating `/profile`-style routes—not when you need a full IdP product comparison. OAuth and session topics appear in metadata; the excerpt emphasizes JWT as the primary path. Pair with your own secret management and rate limiting; the skill teaches structure, not compliance certification. Agents on Claude Code or Cursor can drop these patterns into an existing Fastify app register graph quickly.1installs14Gh Cligh-cli is a comprehensive agent skill that embeds GitHub CLI knowledge for solo builders who live in the terminal and agent-driven workflows. Instead of tab-hopping through docs, you invoke the skill when you need precise flags for creating or reviewing pull requests, triaging issues, watching Actions runs, managing releases, or authenticating against GitHub.com or GitHub Enterprise. The readme structures prerequisites—install via Homebrew, apt, or winget—and walks through gh auth login, token-based login, gh auth setup-git, and routine account switching. Because gh spans coding, shipping, and light operations work, the skill is multi-phase: you will reach for it while integrating git in Build, preparing reviews in Ship, and cutting releases closer to Launch. It is a reference template rather than an autonomous workflow; your agent still executes commands you approve. Intermediate complexity reflects the breadth of subcommands and enterprise auth edge cases. Pair it with repository context (correct remote, logged-in host) before destructive repo or org operations.1installs15Git Commitgit-commit is an agent skill that runs real git operations to produce Conventional Commits. It inspects your diff to infer the right type and optional scope, drafts a clear subject line, and can stage files in sensible batches instead of one giant commit. Solo builders use it throughout Build and Ship whenever they want semantic history without memorizing the spec or writing messages by hand. The skill encodes the full type table—from feat and fix through perf, test, build, and ci—and documents how to flag breaking changes in the footer. Because it is allowed to use Bash, the agent can stage, commit, and adjust in the same workflow you would run locally. It fits Cursor, Claude Code, and other agents that expose shell tools, and pairs well with PR and code-review skills that expect readable commit logs.1installs16Git Flow Branch Creatorgit-flow-branch-creator is an agent skill for developers who want Git Flow naming without debating branch type after every stash. You run one prompt; the agent executes git status, inspects unstaged or staged diffs, classifies the work against the embedded Git Flow Branch Analysis Framework, picks the correct branch type and merge target (for example feature branches from develop), generates a semantic branch name, creates and switches to that branch, then reports analysis and next steps. It reduces mistakes like hotfix names on enhancement work or feature branches cut from main. Solo builders shipping with Copilot, Claude Code, or Cursor benefit when juggling multiple fixes and releases on one repo. Pair it at the start of a change set before commits pile up on main, and again when emergency production fixes need a hotfix path distinct from develop-line features.1installs17Github Issuesgithub-issues is a reference skill for formal GitHub issue dependency management—marking one issue as blocked by another so relationships show in the UI and sync via API. Solo and indie builders shipping with agents often track work in GitHub; when they need blocked-by or blocking edges, native MCP tooling may not cover it, so this skill spells out the REST paths under /issues/{number}/dependencies/blocked_by and the GraphQL fields and mutations that mirror the same model. It is aimed at builders comfortable with tokens and HTTP or GraphQL clients who want repeatable instructions an agent can follow without guessing parameter shapes. Use it during backlog grooming, release planning, or when unblocking a chain of tickets programmatically. It complements general issue-creation skills by focusing solely on dependency edges and the ID versus number distinction that trips up first-time integrators.1installs18Github ReleaseGithub-release in the Harness collection is a commit-classification reference for agents preparing GitHub releases or summarizing what changed since the last tag. Solo builders and small teams use it when automated or human release notes need consistent categories—features, fixes, performance, breaking changes—without misreading marketing-style commit titles. The skill explicitly ranks the diff above the message: heuristics add intent when messages are thin or wrong, not replace reading the patch. It pairs Conventional Commits prefixes with keyword patterns for freeform repos. Intermediate familiarity with git log and release tagging helps. Best invoked as part of a broader harness release flow when the ingested fragment is only the classification table; confirm the full SKILL.md in-repo for release automation steps.1installs19NodeHarness Node is an agent skill package centered on production-minded Node.js HTTP testing for solo builders. The bundled pattern demonstrates how to stand up a minimal JSON API, expose a `/health` probe, and assert behavior across normal operation and graceful shutdown— including correct status codes and connection handling. It targets developers who want copy-paste-ready `node:test` structure instead of improvising async server teardown. Because the readme is implementation-first, treat it as a template skill: your agent adapts the handler and cases to your service. Primary value shows up in Ship when you harden deploy paths, but the same scaffold supports Build when you first wire backend routes. Intermediate complexity reflects async fetch against a live local server and TypeScript-flavored typings in the sample.1installs