AI & ML

waelabouceo Skill

Quickly bootstrap your environment.

waelabouceo Skill

Summarizes a document for the user.

h-mmer Skill

Active vulnerability hunting on a target. Loads scope, reads brain, detects tech stack, runs targeted tests with concrete payloads. Usage: /hunt target.com [--vuln-class idor|xss|ssrf|sqli|ssti|oauth|rce|race|graphql|upload|business-logic|llm-ai]

h-mmer Skill

Build deep exploit chains — dispatches chain-builder agent. Given bug A, recursively walks the chain graph. Usage: /chain (then describe bug A)

h-mmer Skill

Show engagement dashboard with program info, scope, brain state, findings, agent activity, and cost estimate.

h-mmer Skill

Show ranked attack surface for a target. Invokes recon-ranker agent. Usage: /surface target.com

h-mmer Skill

Autonomous hunt orchestrator. INSATIABLE in --autonomous mode: enforces an EXHAUSTION CONTRACT (26 canonical hunter classes, surface probe A-I, depth-engine ≥25 attempts/class, wall-clock floor 90 min/target, PRE-COMPLETION GATE before any summary). No early stops, no clarifying questions, no auxiliary-agent substitution. Usage: /autopilot target.com [--interactive|--autonomous] [--20m-off] [--resume]

immortalqx Bundle

Quick single-paper lookup via AlphaXiv LLM-optimized summaries with tiered source fallback. Use when user says "explain this paper", "summarize paper", pastes an arXiv/AlphaXiv URL, or provides a bare arXiv ID for quick understanding - not for broad literature search.

immortalqx Bundle

Persistent research knowledge base that accumulates papers, ideas, experiments, claims, and their relationships across a project. Inspired by Karpathy's LLM Wiki pattern. Use when user says "知识库", "research wiki", "add paper", "wiki query", "查知识库", or wants to build/query a persistent field map.

speciliam Bundle

Drive a coding task with BOTH Claude and Codex collaborating for better-planned, better-executed output. Invoke as $tandem or /tandem from either the Codex CLI/UI or Claude, in any repository. Whichever model you launch from becomes the driver; it pulls in the other model as its peer to (1) independently critique the plan, (2) optionally split or delegate implementation slices, and (3) review the final diff. Use for non-trivial or long-winded changes where a single model tends to drift, miss edge cases, or under-plan — e.g. 'I want this change in my fantasy-wizard repo and I want both models on it.'

speciliam Bundle

Complete Liam Van's tracked job applications that have tailored resumes but are not yet applied. Use when the user wants one Codex agent to work through `application-trackers/applications.md` rows with `Applied` blank/false and `Status` like `Resume Tailored`, handle live Chrome application flows directly, submit where possible, ask only for blocking form answers or consent-sensitive choices, update the markdown tracker, refresh the recruiting dashboard cache, and perform a light file-based handoff only when context is overloaded.

speciliam Skill

Coordinate Codex and Claude Code as an executor/reviewer pair using local handoff files, Claude subagents, and shared recruiting workflow skills without letting both agents edit the same files at once.

speciliam Bundle

Codex-driven leave-the-PC LinkedIn early-career drain — an interactive Codex (GPT 5.5) session is the conductor AND the browser actor (Codex Chrome plugin first, Computer Use fallback), with best-effort upload-origin warming at a supervised kickoff, park-don't-pause on every manual blocker, isolated /tmp state, and the shared path-parameterized watchdog for caffeinate/RAM/stall/reconcile/notifications. Use when Liam says `$linkedin-unattended-drain-codex`, wants the unattended drain run by Codex instead of Claude, or wants to compare the Codex upload path (absolute local paths, per-origin approvals) against the Claude variant's one-grant model.

bernardorubin Skill

Use when the user wants to hand off the current conversation to a fresh Claude session or another agent — a self-contained handoff document written to the Desktop that a new session reads to pick up exactly where this one left off. Triggers on "handoff", "/handoff", "write a handoff doc", "hand this off to a new session", "pass this to another agent", "context is getting long, write a handoff", "compact this for a new chat to continue". Optional argument describes what the next session will focus on.

bernardorubin Skill

Use when the user wants to scaffold a new Karpathy-style LLM Wiki vault for the current project and register it for the vault-keeper skill. Triggers on phrases like "init a vault", "set up a knowledge vault", "create a second brain for this project", "scaffold the wiki", "vault-init", "start a new vault here".

bernardorubin Skill

Use when working in a project that has a registered second-brain knowledge vault (Karpathy-style LLM Wiki). Auto-fires when the user documents findings (architecture decisions, integration quirks, debugging discoveries, team facts, ticket epics), looks up domain context (team/integrations/flows/past decisions), ingests raw sources into the wiki, or asks for a vault lint/audit. Resolves the current working directory against `~/.config/claude-pro-skills/vaults.json`; if the project has no registered vault, the skill self-terminates without action. To set up a new vault, use `/vault-init`. Triggers on phrases like "save this to the vault", "what does the wiki say about X", "let's document this finding", "ingest this doc", "lint the wiki", "what's our second brain say about Y".

bernardorubin Skill

Use when the user wants a deliberate end-of-session sweep that files everything valuable from the WHOLE conversation into the project's knowledge vault (Karpathy-style LLM Wiki) in one pass — not the ambient single-fact writes that [[vault-keeper]] does during normal work. Reviews the entire session, dedupes against what's already filed, and writes each worth-keeping finding into the right wiki page. Resolves the cwd against `~/.config/claude-pro-skills/vaults.json`; if no vault is registered it says so and points to `/vault-init`. Triggers on phrases like "save to vault", "save this session to the vault", "save whatever's valuable from this session", "dump this session to the wiki", "file everything worth keeping", "/save-to-vault".

bernardorubin Skill

Use when the user wants a change taken from a branch all the way to verified production through the staging ladder. Triggers on "promote to prod", "promote this to production", "push it to prod", "get this to prod", "ship it to production", "promote HPY-1234", "staging then prod", "roll this out to prod". Runs the full ladder — PR into staging, wait for checks to go green, merge, verify the change on staging, PR into main, wait for checks, merge, confirm production is healthy AND that the fix is actually live on prod (via Chrome MCP / playwright-cli / curl), capture evidence of the fix working in production, then comment on each Jira ticket and move it to Done or Ready for QA (asking when it isn't obvious; skipped when there is no ticket). Enforces the promotion discipline: never merge on red or pending checks, never open the main PR until staging proved the change didn't break anything, verify the deployed commit before verifying behavior, and if prod comes back broken, revert immediately. NOT for cutting a

jiangnan030-del Bundle

Orchestrate evidence-backed ITASCA PFC2D/PFC3D studies across planning, modeling, calibration, solving, post-processing, V&V, and delivery; use for full DEM workflows, model audits, or pfc-code-backed case generation.

jiangnan030-del Bundle

Child skill of pfc-workflow for PFC 6.0 contact model selection, CMAT setup, contact properties, bonding methods, and contact-law audits.

jiangnan030-del Bundle

Explain brittle-rock mechanics traits and the limits of the standard bonded-particle model. Use when the user asks about brittle rock formulas, UCS/TS contrast, deep-rock phenomena, BPM assumptions, parallel-bond failure criteria, strength envelopes, or how to organize figures for Chapter 5.1-5.2 style analysis.

jiangnan030-del Bundle

Child skill of pfc-workflow for rapid calibration of an improved PFC3D linear parallel-bond model using 13 micro-parameters, strong/weak contact grouping, Weibull damage, 13-factor orthogonal tests, Pearson correlation, regression formulas, and macro-target back-solving.

jiangnan030-del Bundle

Practical ITASCA PFC 6.0 geotechnical modeling techniques for boundary servo control, particle assembly construction, contact versus cmat assignment, initial-state consistency, loading-rate control, size-effect checks, and FISH-based macro-parameter extraction. Use when the user asks how to build, servo, stabilize, quality-control, or measure a PFC geotechnical model, and especially when they want full PFC 6.0 code rather than a summary.

jiangnan030-del Bundle

Build or explain an equivalent crystal model for brittle rock. Use when the user asks about crystal-network construction, equivalent crystalline modeling, Chapter 5.3 formulas, fine-scale parameter calibration, crack-pattern validation, Hoek-Brown fitting, compression-tension ratio analysis, or how to plot equivalent-crystal simulation outputs.

kuhung Skill

运用《智能体设计模式》(Antonio Gulli) 的 21 个模式,指导 AI Agent 系统的架构设计与选型。适用于用户在设计 Agent 工作流(选择链式/路由/并行/多Agent编排)、提升 Agent 输出质量与可靠性(反思/异常处理/Guardrails)、做 RAG 技术选型(混合检索/GraphRAG/Agentic RAG)、优化 Agent 成本与延迟(动态模型切换/回退)时使用。

canepro Bundle

Use when local Codex MCP or connector setup fails, including Zoho MCP, streamable_http auth, stdio wrappers, mcp-remote, malformed local state, startup warnings, missing features, or add/fix this MCP in Codex tasks. Preserve endpoints the user says work elsewhere; troubleshoot local Codex auth, transport, wrapper, and state first.

pavel-molyanov Skill

Creates, improves, and reviews LLM prompts using concise, task-aware guidance. Use when: "напиши промпт", "улучши промпт", "prompt engineering", "проверь промпт"

pavel-molyanov Skill

Creates, improves, and reviews LLM prompts using concise, task-aware guidance. Use when: "напиши промпт", "улучши промпт", "prompt engineering", "проверь промпт"

pavel-molyanov Bundle

Use when you need information about this project's architecture, tech stack, coding patterns, data model, deployment setup, git workflow, or UX guidelines. Contains comprehensive project documentation including design decisions, technical specifications, and development standards.

tommywagz Bundle

Finds and safely installs existing Agent Skills, MCP servers, or Composio integrations. TRIGGER for "find a skill", "find an MCP server", registry searches, awesome-agent-skills, or Composio integrations. DO NOT TRIGGER for writing a new skill, evaluating an existing skill, or ordinary web/docs research.

tommywagz Bundle

Creates or restructures an Agent Skill package with SKILL.md, scripts, references, and assets. TRIGGER for "create/write/scaffold a skill", a SKILL.md, or fixing a skill's description, token budget, or layout. DO NOT TRIGGER for ordinary documentation, general code, or evaluating a skill.

tommywagz Bundle

Skill for authoring A2A (Agent2Agent) multi-agent workflows, featuring host-agent orchestration and sub-agent communication. TRIGGER when: 'A2A workflow', 'A2A multi-agent', 'A2A host', 'A2A sub-agent', 'Agent2Agent collaboration', 'AgentCard', 'agent-card.json', 'A2ACardResolver'. DO NOT TRIGGER when: normal ADK workflows without A2A (use google-agents-cli-adk-code).

tommywagz Bundle

Evaluates an existing Agent Skill for routing accuracy, task performance, or security. TRIGGER for "evaluate/test/benchmark a skill", trigger precision/recall, red-team/security review, or an eval suite. DO NOT TRIGGER for authoring a new skill, general application tests, or generic statistics.

tommywagz Bundle

Profiles an agent prompt, workflow step, or task spec for cognitive/token complexity, resolves the optimal LLM provider and model tier against a governed catalog, and emits the exact opencode run --model provider/model launch command plus opencode.json agent.<name>.model config patches. TRIGGER when the user asks to 'assign a model for this prompt', 'select the optimal model/tier', 'configure opencode model governance', wants a --model provider/model launch command, needs to reconcile ~/.config/opencode/opencode.json against a project opencode.json, or asks to optimize model choice for cost vs. intelligence (result_maximized, cost_optimized, balanced). DO NOT TRIGGER for rewriting or improving prompt wording (prompt engineering), provisioning API keys/credentials, or running live model benchmarks/evals.

tommywagz Bundle

Searches external Agent Skill, MCP server, and SaaS-tool registries to find an existing skill or integration matching a requested capability, screens every candidate for prompt injection and other security risks before its content is read as instructions, and installs only vetted results into the local project. TRIGGER when the user asks to find, search for, or discover an existing skill, MCP server, or tool integration — e.g. "is there a skill for X", "find an MCP server for Slack", "search awesome-agent-skills", "add a Composio integration for Y". DO NOT TRIGGER for authoring a brand-new skill from scratch with no external source in mind (use a skill-writing skill), scoring/auditing a skill already in this project (use a skill-evaluation skill), or a generic web/documentation search with no skill, MCP-server, or tool-registry angle (e.g. library docs, release notes).

tommywagz Bundle

Builds and audits AP2 (Agent Payments Protocol) payment agents: Checkout and Payment Mandates as SD-JWT verifiable credentials, human-present vs. human-not-present flows, role sub-agents, x402 crypto rails, and authorization/dispute audits. TRIGGER when the user mentions AP2, ap2-protocol.org, 'mandate.checkout.1', 'mandate.payment.1', Checkout Mandate, Payment Mandate, open vs. closed mandate, Credentials Provider or Merchant Payment Processor agents, x402 settlement, google-agentic-commerce/AP2, or asks to let an agent pay, delegate autonomous spend under signed constraints, or prove after the fact who authorized a purchase. DO NOT TRIGGER for UCP cart/catalog/checkout-session work with no payment settlement (use ucp-merchant-servers or ucp-consumer-surface), or for wiring a non-agentic payment gateway SDK (Stripe, PayPal, Adyen) with no verifiable mandate credential.