03 · Skill Guides
Start with our latest recommendations. The archive shows dated guides and clearly labeled source-available dates for retrospective picks.
Start here
Practical workflows from the latest guides and skill updates.
Guide a coding agent through dependency and secrets checks, cross-file data-flow analysis, self-verification and a structured security report. It proposes patches for critical and high findings and requires human review before applying them.
CodingView skill
Two Skills help an AI agent inventory PHP patch status without exploit probes. phpcheck-url reads advertised versions from HTTP headers, the labels sent with a web response; phpcheck-server directs the agent to verify the actual PHP-FPM or Apache build and its package advisory. The bundled checker can return OK, VULNERABLE or UNKNOWN and emit JSON.
CodingView skill
Guides an AI coding agent through Java 11→17, 17→21 or 21→25 migrations. It reads Maven, Gradle and CI configuration, selects the relevant version guide, updates the build before adopting language features, and calls for builds and tests at each step. The October 15 change replaces large always-on instruction files with an on-demand Skill and version-specific references.
CodingView skillExplore past recommendations
Browse past guides and updates by month.
Dates identify publication or directory addition, except rows labeled “Source available by.” Those retrospective dates mark when the cited version already existed; the directory shows when TechReadly added and reviewed it.
Find a tool for the job
31 skills, organized by what you want to do.
Create, revise, and inspect Word documents with professional formatting.
Build and analyze spreadsheets while preserving formulas and layout.
Create and edit slide decks with reusable layouts and visual checks.
Read, generate, transform, and visually verify PDF documents.
Apply controlled-language guidance to technical instructions.
Guide a coding agent through dependency and secrets checks, cross-file data-flow analysis, self-verification and a structured security report. It proposes patches for critical and high findings and requires human review before applying them.
Two Skills help an AI agent inventory PHP patch status without exploit probes. phpcheck-url reads advertised versions from HTTP headers, the labels sent with a web response; phpcheck-server directs the agent to verify the actual PHP-FPM or Apache build and its package advisory. The bundled checker can return OK, VULNERABLE or UNKNOWN and emit JSON.
Guides an AI coding agent through Java 11→17, 17→21 or 21→25 migrations. It reads Maven, Gradle and CI configuration, selects the relevant version guide, updates the build before adopting language features, and calls for builds and tests at each step. The October 15 change replaces large always-on instruction files with an on-demand Skill and version-specific references.
The `batch-script-progress` Skill provides a specification for adding startup notices, bounded progress updates, sanitized per-item failure diagnostics, and terminal summaries to automation scripts that iterate through records, files, or API endpoints. Its stated design goal is to deliver runtime visibility without altering the existing processing semantics, exit statuses, retry boundaries, or stdout data contracts of the underlying code. The specification preserves stdout for its established result or protocol output, defaulting diagnostic messages to stderr unless callers reject that channel, in which case developers must use an existing logger or halt if no safe stream is available. In its accounting model, the Skill enforces that unknown totals remain explicitly unknown when streams are lazily evaluated, strictly barring the invention of artificial ETAs, completion percentages, or premature stream materialization. Terminal counts follow an accounting rule where processed items equal the sum of succeeded, failed, and skipped items, counting each unit once regardless of internal retry attempts. In-flight tasks cannot be marked successful before durable commits occur, rolled-back transactions must be logged as rolled back rather than completed, and scripts that traditionally return exit code zero during partial failures must preserve that exact exit status while emitting an honest terminal summary.
The python-mcp-server-generator skill instructs GitHub Copilot Agent Skills hosts to scaffold, implement, and configure a Python-based Model Context Protocol (MCP) server that exposes local Python functions, existing APIs, or databases as structured tools for AI clients. The skill generates an application project layout using uv with --no-package, installs the mcp[cli]>=2,<3 package alongside pytest, and writes a standardized server.py using typed schemas and docstrings to automatically generate tool input definitions. It configures stdio as the primary local transport, supports optional Streamable HTTP references, establishes in-memory Client integration tests via anyio without requiring subprocesses or network ports, and outputs client connection manifests for Visual Studio Code (.vscode/mcp.json) and Claude Desktop (uv run mcp install server.py).
Connects a CopilotKit React or Next.js frontend to a Microsoft Agent Framework agent deployed as a Microsoft Foundry hosted agent through an intermediary AG-UI server gateway. The revised workflow uses native Foundry conversations, durable tool approvals and background execution. Hosted agents do not speak AG-UI directly; the default gateway connects to their Responses endpoint and keeps Microsoft Entra ID authentication on the server.
Configuring, provisioning, and validating Claude Platform Managed Agents from bundled console templates or remote documentation URLs.
Prepare Windows desktop MSIX submissions for Microsoft Partner Center: capture product identity, inspect packages, run the Windows App Certification Kit, and assemble listing and certification notes. Binding agreements and certification submissions require action-specific approval; the skill does not guarantee certification.
Guide reviewers through submission standards for contributions to github/awesome-copilot, including fit, value, provenance and AI disclosure, differentiation, supporting evidence, marketing claims, and maintainability alongside deterministic checks.
Compare C# API contracts with JavaScript or TypeScript consumers to find likely breaking changes.
Inspect failing GitHub Actions pull-request checks and their logs before planning a fix.
Compare documentation with code, configuration, and API behavior.
Review existing assertions and identify missing regression test scenarios.
Inspect and test local web applications through browser automation.
Design MCP servers that expose reliable tools to AI agents.
Give coding agents structured workflows for specification, planning, implementation, testing, review, and shipping.
Generate spatial SQL and inspect results on a map.
Load shared team rules and decisions when coding tasks need them.
Combine requirements interviews, test-driven work and code review.
Ground web or iOS interface work in the product brief, existing design system, and one focused playbook.
Generate hand-drawn architecture, flow, and UX diagrams.
Bring Apple HIG, Material Design 3, and WCAG 2.2 guidance into product work.
Turn product direction into distinctive, production-oriented interfaces.
Compose original visual assets as polished PNG or PDF files.
Turn video into frames and transcripts an agent can inspect.
Produce compact animated GIFs tuned for Slack delivery.
Check the source repository for installation steps, requirements and compatibility.