Start with the job, then find an open-source answer

grayscale converts one image, many images, a directory, a ZIP archive, or any mix of them through a Claude skill or a plain Python command. It adds a -Grayscale suffix, preserves alpha channels for PNG, WebP, GIF, and TIFF, and keeps folder structure when rebuilding an archive while removing macOS junk and non-image files. Pillow is the only library dependency, with options for recursive folders and a custom output directory.

Web Experience Auditor splits a website or web-app audit across specialized subagents for scope, browser capture, layout, UX, KPIs, structured data, research, and synthesis. Its knowledge gate checks installed versions before making version-specific recommendations and can fall back to static fetching when browser capture is unavailable. Full runs need a Claude Agent SDK environment and optional browser tooling.

Forced-Holding Screen uses de-identified tickers and descriptions to map where index funds may force exposure to a stock, classify the risk, and show category-level alternatives that keep similar market exposure. It checks current IPO status and fund holdings at run time, so the result can change. The project is informational, not investment, tax, or legal advice, and should be checked against primary sources.

Gov Writing Skill turns notes or rough drafts into formal Chinese government and Party documents, including reports, summaries, speeches, implementation plans, minutes, letters, and Party-building material. It can switch between strategic provincial language and more operational county or district wording, while applying structured political framing and parallel prose. It is a drafting and editing aid, not a source of official facts.

ADHD Filter restructures notes, documents, or pasted conversations for easier scanning while preserving every fact, equation, definition, link, and nuance. It can add a bottom-line summary, callouts, chunking, a table of contents, and answerable reflection prompts, then checks element counts before writing. Files marked in frontmatter are skipped on later runs unless forced.

P2H2P converts an old deck, rough idea, whiteboard photo, or screenshot into editable HTML and PPTX. It orchestrates template selection, parallel HTML building, visual checks, conversion, and final QA, pausing at outline, template, and HTML-preview checkpoints. It is an orchestration layer rather than a standalone renderer and needs the companion HTML-template and PPTX skills plus their runtime tools.

Shizheng aggregates CCTV’s Xinwen Lianbo and domestic-news sources into a daily current-affairs brief. It supports date and source selection, full or brief output, local deduplication, configurable limits, and Markdown archiving without sending the full text through the model. It needs Node.js 18+ but no API key; source availability, freshness, and the rights to republish content remain outside the skill’s control.

PITH compresses long natural-language handoffs between agents with local word-information scoring, while quarantining code, JSON, URLs, and paths. A negation check protects polarity, and a bounded sentence-length gate can reduce the pruning ratio and retry. Payloads below 10,000 characters pass through unchanged; the skill, MCP server, and Python CLI work offline without external APIs.

BA Project Factory gives business analysts 24 connected prompts spanning initiation, discovery, planning, delivery, testing, and release. Each prompt generates an artifact, asks for user approval before storing it, and flags related documents that may need updates. Outputs include Confluence-ready Markdown, Jira-ready stories, Figma-ready wireframe specifications, and meeting or requirements registers, even when source material is incomplete.

Zero Docs mirrors Rocicorp Zero documentation into version-pinned, grep-friendly Markdown so Claude Code can answer ZQL, schema, mutator, permission, and zero-cache questions without a network round trip. A sync script refreshes the local files and records the upstream commit, with a weekly action for updates. It is an unofficial community mirror and needs maintenance when the source changes.

s5cmd Skill teaches Claude Code to use s5cmd v2.3.0 for bulk S3-compatible transfers instead of slow command loops. It covers object copy, sync, deletion, streaming, metadata, presigned URLs, and bucket operations, with notes on quoting wildcards and previewing deletions. The reference includes MinIO, R2, and other endpoints, but the s5cmd binary must be installed separately.

GHL Claude Skill lets Claude Code operate GoHighLevel through API v2, covering contacts, opportunities, custom fields, tags, lookups, and spreadsheet imports with in-memory deduplication. It handles the service’s browser User-Agent requirement, E.164 phone normalization, padded spreadsheet rows, trusted names, and one opportunity per funnel. Credentials stay in environment variables, and the skill distinguishes API-supported objects from UI-only features.

Second Draft revises prose by diagnosing five causes of an AI-like voice: missing stakes, performed emphasis, excessive symmetry, hidden actors, and abstraction. Its four-pass method uses frequency budgets instead of banned-word lists and changes rules by register, so technical references do not receive essay-style edits. Six falsifiable yes-or-no checks keep the revision tied to concrete consequences rather than a self-assigned score.

ClaudeGuard is a report-only pre-merge policy gate. It reads a branch diff, selects matching core and project rules, evaluates added lines, and emits PASS, WARN, or FAIL plus parseable JSON. Version 0.2.0 adds reproducible candidate detection and exemptions for selected rules, fixture tests without an API key, and a runner verdict line. The gate never edits, stages, or commits code; CI maps FAIL to a blocking exit.

Job Search Ops coordinates a five-phase job search: document intake, a structured career-profile interview, a master CV, per-posting fit and tailored PDF documents, then interview coaching with a printable cheatsheet. It keeps applications consistent through a database record, scores capability and preference separately, and preserves honest gap wording. It never fabricates experience, auto-applies, scrapes job boards, or sends applications.

Myra Agents skills is a small catalog of installable skills for common agent workflows. It includes a searchable code index, runtime debugging with direct logs, launch-content drafting from changelogs, persistent-memory consolidation, and accessibility-tree tests for native macOS app surfaces. Each skill is a self-contained folder that works with compatible agents such as Claude Code, OpenCode, Codex, and Cursor, and can be installed individually or as a set.

AI SEO Maxer helps audit and improve a website’s chances of appearing in Google AI Overviews, AI Mode, and assistant citations. Its workflows check indexability, snippet eligibility, content quality, search intent, technical foundations, first-hand value, and agent access, then produce prioritized fixes or citation-oriented rewrites. A separate mythbusting mode tests GEO and AEO advice against Google’s published guidance instead of treating every trend as a required tactic.

set-the-mood changes a Claude Code session’s tone through three combined behaviors: terse phrasing, vivid plain-language analogies, and one-question-at-a-time interrogation of plans against project context and architecture decisions. It can update CONTEXT.md and offer an ADR as decisions settle, while keeping full explanations for security warnings, irreversible actions, and tricky sequences. The mood starts at the beginning of a chat and stays active until reset.