dust-breaking-changes
CRITICAL guideline - Never introduce breaking changes to the private API without explicit user approval. Always warn and ask for validation first.
apply-fix
Apply an approved Obelus plan file to the paper source.
formax-approval-ui-workflow
Use when adding or changing any Approval prompt/presenter.
geo-audit
GEO audit and autofix tool. Audits any website for AI search engine optimization (ChatGPT, Claude, Perplexity, Gemini, Google AI Overviews) and traditional SEO. Generates fix files automatically and recommends code-level changes for approval. Use when user says "geo", "seo", "audit", "AI search", "AI visibility", "citability", or provides a URL for analysis.
codex-review-loop
Adversarial PR code review using Codex CLI. Codex review (~20-50 min) -> structured findings -> HITL approval -> fix loop.
Lassare
Human-in-the-loop for AI coding agents — ask questions, get approvals via Slack.
cactus
Ask the human without stopping work. Use cactus instead of AskUserQuestion whenever a question can be posted now and collected later, whenever several decisions belong to one thread, whenever you are announcing what you will do anyway and only need a veto, whenever a command needs approval, or whenever a verify block or an ordered checklist should stay on the human's board across sessions. Triggers on "cactus", "ask the human", "park this question", "put it on the board", "I need a decision but keep going". Do NOT use for a question whose answer you need before the next keystroke and that has no sensible default; that is AskUserQuestion.
enrichment-module-builder
Build a new Nemesis file enrichment module end-to-end with explicit user approval gates for output mode, library choice, sample files, and integration testing.
legax
Use when the user wants an agent session bridged to a phone, wants remote replies polled, or wants the agent to ask a phone for approval before sensitive actions.
component-consistency-audit
Library-wide audit of how components treat the same thing. Finds controls, marks and surfaces drawn more than one way, components that redraw a shared primitive instead of composing it, and surface families that split on fill, hover, radius, selected or disabled, then fixes the defects for approval. Use when asked to audit consistency, check whether components look and feel alike, find redrawn controls or duplicated primitives, or review surface treatments across the library.
Approval Studio
Approval Studio is an online proofing software for creative teams and everyone working with designs
codex-code-reviewer
Systematic code review workflow using zen mcp's codex tool. Use this skill when the user explicitly requests "use codex to check the code", "check if the recently generated code has any issues", or "check the code after each generation". The skill performs iterative review cycles - checking code quality, presenting issues to the user for approval, applying fixes, and re-checking until no issues remain or maximum iterations (5) are reached.
Io.Github.Tponscr Debug/Oracle H
Human approval gate for AI agents. Critical actions approved or rejected by a human via Telegram.
backpack-code-review-checklist
Comprehensive code review checklist for Backpack design system components. Use when: (1) Reviewing PRs for new or modified Backpack components, (2) Validating component compliance with Constitution and design system rules, (3) Checking if component follows Backpack conventions before merge, (4) Identifying violations in API design, token usage, accessibility, or documentation. Covers Constitution principles (I-XIII), decisions/ guidelines, API encapsulation rules, private token restrictions, design approval workflow, icon alignment helpers, hover mixin usage, token semantic correctness, and snapshot currency. Essential for maintaining Backpack quality standards and catching non-obvious violations like className props in new components, wrong icon alignment wrapper, raw :hover instead of bpk-hover mixin, or cross-component private token usage.
Autopilot â Idea to Merged PR Pipeline
Drives the full flow from raw user idea (or an existing spec) through merged PR. The ONLY pause is explicit user approval of the spec after Step 0; everything after that runs unattended unless blocked by an unrecoverable failure, missing credentials, or an unresolved CRITICAL Codex finding.
Design reference
Structured design research process that builds visual direction from reference review and stakeholder feedback. Produces a `DESIGN_RESEARCH.md` document and an approved prototype.
ql-brainstorm
Part of the quantum-loop autonomous development pipeline (brainstorm → spec → plan → execute → review → verify). Deep Socratic exploration of a feature idea before implementation. Asks questions one at a time, proposes 2-3 alternative approaches with trade-offs, presents design section-by-section for approval, and saves an approved design document. Use when starting a new feature, exploring an idea, or before writing a spec. Triggers on: brainstorm, explore idea, design this, think through, ql-brainstorm.
github-issue-solver
Accept a GitHub issue URL or issue details, create a comprehensive plan (in plan mode - shows plan first), then implement the solution including tests and documentation, create/update release notes for Home Assistant users, and update the README roadmap for new features. Use when the user provides a GitHub issue or asks to solve an issue. ALWAYS presents the plan first and waits for user approval before implementing.
Agentforge
DeFi safety layer for AI agents: wallet checks, contract docs, approvals, tx decode.
/agent-governance â Agent Capability Management
Manage which agents can do what, with approval gates for adding/removing capabilities. From Paperclip's governance framework.
FDA Data MCP
Clean FDA regulatory data: company resolution, facilities, recalls, inspections, approvals.
the-loop
"the-loop" is an opinionated product-development-lifecycle (PDLC) harness, shipped as a plugin for Claude Code, Cursor and Codex. Once a work item's spec chain (requirements â design â testing plan â tasks) is approved, the harness executes it end-to-end with MINIMAL or NO human intervention,
backlog-prioritize
Fetch all approved-for-grackle issues, filter out closed and already-in-progress ones, and suggest a prioritized ordering. Run with /backlog-prioritize.
chrome-cdp
Interact with local Chrome browser session (only on explicit user approval after being asked to inspect, debug, or interact with a page open in Chrome)
Claude Buddy
The user runs Claude on their Mac (Claude Desktop and/or Claude Code). A companion daemon on this device bridges them: it surfaces Claude's state and asks you to approve tool permission prompts out loud, so the user can say "yes" instead of clicking. The daemon listens on `127.0.0.1:5002`.
architecture-lead
Architecture consultation for Conductor orchestrator. Makes decisions about system design, patterns, component boundaries, and technical architecture. Can approve architectural choices within established patterns. Escalates novel patterns or breaking changes to Board of Directors.
approval-development
Build and work with cjmellor/approval features, including approval workflows, custom states, time-based expirations, rollbacks, and event handling.
mykg
Run mykg knowledge-graph commands inside Claude Code from one slash command `/mykg`. The user describes intent in natural language (extract, append, resume, approve, walkthrough, parse-docs, fetch-web, query); the skill parses intent, builds the right `mykg` CLI command from the live `--help` output, confirms, runs it, and drives the inbox/outbox watch loop for LLM-bearing commands (extract-graph). For read-only queries, prefers MCP tools when the mykg MCP server is online, falling back to reading session files directly. Ensures `.mcp.json` is configured (with user approval). Excludes `mykg init` (interactive shell command) and `mykg merge-graphs` (follow-up planning).
agent-safety
Use when bounding an LLM agent that already runs â scoping its task domain, gating tools to least privilege, defending against prompt injection in untrusted web/email/RAG text, requiring human approval on irreversible actions, capping runtime and cost, or triaging what it already did. NOT building the loop, tools, or RAG (that is `building-agents`).
linkedin-comment-drafter
Draft a LinkedIn comment on someone else's post from its URL. Use when the user pastes a post URL and asks to comment, engage, or be first commenter. Produces 1-3 variants in the user's voice, picks a reaction, and schedules via Publora on approval. Not for replying to existing comments (use linkedin-reply-handler).
Aegis — AI Agent Governance
Policy-based governance for AI agent tool calls. YAML policy, approval gates, audit logging.
aident-loadout
Coordinate multi-step work across services the user has authorized through Aident Loadout, using live action schemas, connection checks, cost preflight, and explicit approval for side effects.
Decern
MCP server for Decern CRM — contacts, deals, pipelines, tasks, and approvals.
OpenTelemetry Instrumentation Extension
Extend OpenTelemetry instrumentation when new functionality is added to the MCP Gateway. Use when (1) new operations/functions are added, (2) reviewing code for missing instrumentation, (3) user requests otel/telemetry additions, or (4) working with state-changing operations. Analyzes git diff, suggests instrumentation points following project standards in docs/telemetry/README.md, implements with approval, writes tests, updates documentation, and verifies with debug logging and docker logs.
Io.Github.Vish288/Mcp Gitlab
MCP server for GitLab API — projects, MRs, pipelines, CI/CD, approvals, issues, code review.
approval
设置 omi æ¹å姿æï¼`suggest`ã`auto`ã`full-auto`ï¼å¹¶å¸¦ææç¡®çé£é©æ¤æ ã
implement-change
Executes plan.md for an active change under Docs/<scope>/Specs/<change-name>/. Use after plan-change when the user approves implementation. Follows tasks in order, validates each step, retries obvious failures once, and stops to ask if blocked. Does not change spec intent without approval.
Payclaw
Virtual Visa cards for AI agents. JIT card issuance, human-approved, MCP-native.
backlog-technical-project-manager
Technical project management for Backlog.md workstreams using coordinated sub-agents. Use only when the user explicitly asks Codex to take over one or more existing Backlog.md tasks (for example: "act as TPM", "coordinate these tasks", "delegate to sub-agents"). Orchestrate planning, implementation, and finalization; manage dependencies and overlapping code areas; review and approve sub-agent plans before implementation.
character-sprite
Generate complete character sprite sheets for the Claude Office Visualizer agents. Creates all animation frames (idle, walking, typing, handoff, coffee) with consistent character design across all sheets. Uses iterative approval workflow and reference-based generation for consistency.
Packet Factory
The packet factory is the gatekeeper between intent and execution. Nothing with a side effect runs without an APPROVED packet.
tuios
Drive tuios from inside one of its panes. Find out where you are running, read and write other panes, run work and wait on it instead of polling, report your own state so the person sees it, and talk to the other agents and the person safely. `tuios --skill <topic>` prints the rest: fleets of agents, turn checkpoints, shipping a worktree, the Inbox and approvals, push notifications, mail, other machines, clients, events, MCP, the tmux shim, herdr's command line, pane grants, configuration, agent features off, errors and recipes.
agentic-sdd-architecture-review
Review an SDD implementation plan before code generation and decide whether it is approved, needs changes, or is rejected.
Io.Github.Kar69 96/Agentpay
AI agent payments with human approval. Local-first, encrypted credentials, budget controls.
engram
Local-first personal AI identity and memory layer for MCP-compatible coding tools (Claude Code, Codex, Cursor, and others). Use this skill when the user wants to continue from a previous session ("continue from last session", "pick up where we left off"), recall a past decision ("what did we decide", "what was our reasoning"), persist something durable ("remember this", "save a lesson", "save a decision", "save a playbook"), search prior knowledge ("search what we know about X", "have we hit this before"), export their identity or context ("export my identity card", "give me my context"), or maintain local-first cross-tool identity and memory that the user owns and approves. Engram stores user-approved lessons, decisions, playbooks, and project context as local JSON; the AI suggests, the user decides what becomes permanent.
SiliconBridge
Human-in-the-loop API for AI agents. CAPTCHA, OTP, KYC, and approvals by real humans.
saguaro-createrule
Generate a single Saguaro review rule using the AI pipeline with preview and approval
Clawvault Mcp Server
AI agent payment security - spending limits, whitelists, and human approval.
claude-handoff
Orchestrate read-only Explore research subagents during planning and one to eight Sonnet subagents to implement the approved plan.
agents-md-improver
Audit and improve project-rules files (AGENTS.md, CLAUDE.md, .agents/instructions, local overrides) so the agent keeps accurate project context. Use when the user asks to check, audit, review, update, improve, or fix their AGENTS.md or CLAUDE.md, mentions "project rules maintenance" or "agent context optimization", or when the codebase has changed enough that the rules file may be stale. Scans the repository for every rules file, grades each against a quality rubric, outputs a quality report, and applies targeted edits only after user approval.