Claude Skills
The open format is called Agent Skills and works in Claude Code, Codex, Cursor and other agents — most people know it as Claude Skills.
Every Agent Skill we could find on GitHub, deduplicated by content. 79 870 files from 1 769 authors, of which 62 217 are unique — the rest is the same skill repackaged into someone else's repository. For each one: what it weighs in tokens, whether it ships runnable scripts, and which MCP servers it needs.
Checkable-ish surface presentation signals a reviewer notices first — duplicate/near-identical tables, leftover pipeline/template strings, too-few or LLM-looking figures, and page-padding. AUXILIARY ONLY and weak by design: a deterministic pass (tools/check_presentation.py — dup-table + pipeline-artifact) plus a fresh cross-model GROSS-cases-only semantic pass (thin-float, LLM-figure, page-padding), every above-info finding span-anchored to the evidence ledger (claims.json). The adjudicator CAPS everything at minor (SURFACE_ONLY_SKILLS + SURFACE_PATTERNS) — these contribute at most SOFT_FLAGS, never a HARD verdict — default false_positive_risk:high. NOTE: the pure AI writing-STYLE impressions (AI-flavor prose, defensive 'not-X-but-Y' hedging, narrative-arc, jargon-stuffing, invented codenames) MOVED to the zero-verdict-weight AIS track — for those use skills/ai-style-impressions, NOT this. Emits presentation-signals.findings.json; NEVER computes the verdict. Triggers: \"presentation signals\", \"surface check\", \"duplicate tables\", \"排版信号\".
End-to-end substantive-integrity forensic sweep of a research paper (especially autoresearch / AI-Scientist-style output). Orchestrates the whole pipeline: ingest (arxiv-id | pdf | dir → working dir + pdftotext for L0) → /evidence-ledger (artifact manifest + observability level L0/L1/L2 + span-anchored claims.json) → fan out the integrity auditor skills (consistency, citation, baseline, experiment, presentation, proof-derivation, eval-design — each reads the ledger, emits span-anchored findings) + the zero-verdict-weight AIS writing-style track → advisory memos (/adversarial-case-builder + /novelty-duplication-advisory, no verdict weight) → deterministic tools/adjudicate_findings.py (--ledger REQUIRED) → reviewer-ready Integrity Forensics Report. Cross-model (fresh codex per dimension) and reviewer≠adjudicator: the model proposes findings, the deterministic adjudicator decides the verdict. Observability-aware, detect-only, never an opaque AI-text classifier (a separate zero-weight AIS section lists AI writing-style impressions, never moving the verdict). Triggers: \"anti-autoresearch\", \"integrity audit this paper\", \"forensic review\", \"audit a submission\", \"审一篇投稿的诚信\".
Cognitive triage of fleet audit findings. Read issue content, evaluate labeling accuracy, and determine open/close/dispatch/relabel actions for fleet-managed issues. Use when triaging undispatched issues or reviewing audit scan results.
Comprehensive instructions for executing tasks using the gws (Google Workspace) CLI or similar agent-first command-line tools. Use this skill when interacting with machine-readable CLIs to ensure safe mutations, enforce context window discipline, and avoid input hallucinations.
Recall and write source-attributed local memory.
Write deliberate source-attributed local memory.
Set up machine-wide agent memory and open the local project workbench.
Set up machine-wide memory and open the local project workbench.
Deliberately write or revise one complete curated memory file.
Recall source-attributed local memory before work.
Work from the same local memory as every other coding agent on this machine. Recall prior decisions, preferences, instructions, and project conventions that Claude Code, Codex, Cursor, and other agents wrote here, with cited sources, fully local. Also searches a separate local technical-documentation index.
Inspect Docmancer tree, index, capture, and security health.
Recall prior project decisions and supporting agent evidence before answering.
Relight an existing JPG, JPEG, PNG, or WebP portrait and rebuild clean photographic skin response without changing the person. Use for realistic skin recovery, plastic-skin removal, natural fill, backlight correction, soft window light or shadows, tree shadows, bokeh, golden-hour side backlight, neon, studio soft light, low-key light beams, full-black silhouettes, or Higgsfield-Relight-like image edits.
Generates fast, SEO-optimized static HTML landing pages targeting 100/100 PageSpeed (LCP < 2.5s, INP < 100ms, CLS < 0.1), full schema.org JSON-LD, AVIF images, critical CSS, zero external dependencies. Use when: user asks to create/build/generate a landing page, one-pager, or static site with focus on SEO, speed, or PageSpeed; asks for an SEO-friendly page from a brief/ТЗ; or asks to audit/fix a landing against a performance checklist.
Migrating CI/CD pipelines to TeamCity. Use when the user wants to migrate, convert, or switch to TeamCity from GitHub Actions (.github/workflows/) or Bamboo (bamboo-specs/*.yml), even if they only say "move our CI". Other CI systems (GitLab, Jenkins, CircleCI, Azure DevOps, Travis, Bitbucket) are not supported yet.
Use when working with TeamCity CI/CD or when a user provides a TeamCity build URL — drives the `teamcity` CLI for builds, logs, jobs, queues, agents, pools, projects, and pipelines.
Use for BootstrapFewShot, bootstrapped demonstrations, teacher-model demos, and low-data DSPy prompt optimization.
Use for creating custom DSPy modules, extending dspy.Module, reusable components, stateful modules, serialization, and module testing.
Use for BetterTogether, prompt plus weight optimization, fine-tuning sequences, and strategy chains like p -> w -> p.
Use for debugging DSPy programs, inspect_history, tracing LLM calls, custom callbacks, observability, monitoring, and cost tracking.
Use this skill when you need to QA audit and fix a plugin skill file. Provides a methodology for verifying skill content against official documentation, fixing issues in-place, and producing verification reports.
Use for composing DSPy modules with Ensemble, MultiChainComparison, ensemble voting, sequential pipelines, and multi-program workflows.
Use for DSPy adapter selection, JSONAdapter, XMLAdapter, ChatAdapter, native function calling, structured outputs, and multimodal inputs like dspy.Image or dspy.Audio.
Use for DSPy retrieval with dspy.Embedder, dspy.Embeddings, FAISS indexes, semantic search, and local or hosted embedding models.
Use for BootstrapFinetune, fine-tuning DSPy models, teacher-student distillation, weight optimization, and lower-cost deployment.
Use for integrating DSPy with Haystack, optimizing Haystack prompts, improving retrieval pipelines, and extracting DSPy prompts.
Use for GEPA reflective optimization, ReAct agent optimization, feedback metrics, LLM reflection, and execution trajectories.
Use for evaluating DSPy programs with Evaluate, answer_exact_match, SemanticF1, custom metrics, baselines, and program comparisons.
Use for MCP tools with DSPy, Model Context Protocol servers, dspy.Tool.from_mcp_tool, and ReAct agents over MCP-compatible tools.
Use for MIPROv2, Bayesian optimization, instruction and demo tuning, and high-performance DSPy program optimization.
Use for GEPA optimize_anything on text artifacts such as code, prompts, agent architectures, configs, and non-DSPy optimization targets.
Use to choose or compare DSPy optimizers including LabeledFewShot, BootstrapFewShot, MIPROv2, SIMBA, GEPA, BootstrapFinetune, Ensemble, and BetterTogether.
Use for dspy.Refine, dspy.BestOfN, output constraints, validation, reward functions, and iterative output refinement.
Use for deploying DSPy with save/load, configure_cache, restrict_pickle, track_usage, async execution, streaming, and production runtime controls.
Use for ReAct agents, tool-calling agents, dspy.ReAct, multi-step reasoning and acting, and GEPA agent optimization.
Use for DSPy reasoning modules including RLM, ProgramOfThought, CodeAct, Parallel, sandboxed execution, and long-context workflows.
Use for DSPy signatures, InputField, OutputField, typed inputs and outputs, signature classes, and Pydantic-style structured schemas.
Use for RAG pipelines, retrieval augmented generation, ColBERTv2, context retrieval, multi-hop RAG, and grounded DSPy answers.
Use for SIMBA optimization, mini-batch introspective optimization, self-reflective rules, stochastic ascent, and numeric-metric optimization.
Use when a Hermes profile must be audited for role clarity, authority boundaries, configuration fit, skills, memory posture, credential scope, handoffs, and recurring operational failures.
Use when installed Hermes skills must be audited for overlap, staleness, broken references, usage-integrity problems, and dead weight without changing the installation.
Use when Hermes messaging gateway failures must be diagnosed across process state, adapters, credential posture, logs, delivery evidence, polling conflicts, and service persistence without automatic repair.
Use when a top-level, read-only Hermes health audit is needed across installation, updates, gateways, cron, profiles, skills, repositories, credential posture, persistence, and cost signals.
Use when stuck, contradictory, blocked, or incomplete Hermes updates must be diagnosed across Git state, remotes, running processes, caches, and installed versions before recovery.
Use when a Hermes environment must be safely migrated between machines with staged exports, integrity manifests, secret separation, selective imports, verification, and rollback.
Use when the user says Interview me before you start as a standalone command, explicitly asks to be questioned before work begins, or needs an adaptive, consent-based interview because goals, constraints, preferences, tradeoffs, or success criteria are genuinely unclear.
Use when Hermes token usage, cost attribution, runaway sessions, cron consumption, or billing discrepancies must be investigated using privacy-preserving, schema-aware evidence.
Use when a proposed open-source feature must be checked across source, history, branches, issues, pull requests, roadmaps, and contributor guidance before implementation begins.
Use when a user asks whether an identified repository is ready for further development, release work, a new feature, handoff, or a new contributor, requiring a disciplined read-only audit before an evidence-backed verdict.
Use when an open-source developer tool, package, CLI, agent, or MCP server must be evaluated for legitimacy, supply-chain risk, telemetry, dangerous capabilities, claim accuracy, and adoption fit.
Use when X Analytics CSV exports must be inspected, validated, normalized, imported, or compared through a repeatable private-by-default workflow.
Use when drafting, rewriting, or repurposing short-form X content, including single posts, quote posts, replies, threads, launches, and personal stories, with source fidelity and claim verification built in.
Create, revise, and audit executive-ready strategy-consulting presentations as native, editable PowerPoint files. Use when turning topics, text, Markdown, Word, PDF, spreadsheets, CSV data, URLs, research, or an existing PPTX into a board, client, strategy, transformation, market-analysis, or McKinsey-style consulting deck; when applying a brand or PowerPoint template; or when checking a deck's storyline, evidence, charts, layout, and visual quality.
Turn Chinese classical poems and ci into coherent vertical Chinese-art videos with poem-driven scene grouping, GPT ImageGen stills, Docker-only Gemini I2V, retained model-generated ambience, Gemini sparkle-watermark cleanup, brush-calligraphy captions revealed character by character, optional local BGM mixing, stitching, and final-frame QA. Supports verse-driven variation across ink landscape, gongbi bird-and-flower, colored figure-and-horse painting, blue-green landscape, xuan paper, silk, and related Chinese visual languages. Use when users ask for 古诗词动态视频、诗词逐句或两句一景、国风视频、毛笔字逐字出现、整首诗拼接成片，or want the established 月落乌啼霜满天 workflow applied to another poem.
Debug GitHub Actions workflows by downloading logs, analyzing summaries, and understanding how agentic workflows and the AWF firewall work together.
Prepare an open pull request for merge from a GitHub Copilot cloud agent. Drives Reviews, local validation, and Mergeable to a ready state. Does not merge, and cannot trigger CI.
Google C++ Style Guide rules for writing clean, maintainable C++ code. Use when writing C++, reviewing code, discussing naming conventions, formatting, class design, or any C++ best practices. Covers headers, scoping, classes, functions, naming, comments, and formatting.
Check C++ naming conventions against Google Style Guide. Use when checking if names follow conventions or when renaming identifiers.
Claude Skills — questions
Answers built from the skills we actually parsed.