M2AI Skills Pack
m2ai-portfolio/m2ai-skills-pack
185 skills portáveis para construir, governar e operar agentes IA
Ver no GitHubColeção de 185 skills portáveis para Claude Code organizadas em 7 plugins temáticos. Cobre arquitetura de agentes, segurança, otimização de custos, pipelines de código e análise de negócios. Instale apenas os plugins que precisa.
Instalação
npx convez-ai m2ai-portfolio/m2ai-skills-pack1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Arquitetura e design de sistemas multi-agente
- Implementação de guardrails e verificação de segurança
- Otimização de custos e roteamento de modelos IA
- Governança e operação de agentes em produção
- Automação de análise de código e documentação de APIs
Veredito

Pack robusto e maduro para times que operam agentes Claude em escala. Foco forte em governança, custos e safety — essencial para quem precisa de agentes confiáveis em produção. A divisão em 7 plugins independentes permite adoção modular.
README
A tradução para o português deste README ainda está na fila.
M2AI Skills Pack
185 portable Claude Code skills for building, governing, and operating AI agents: architecture, safety, model strategy, cost control, code pipelines, and business analysis. Ships as 7 themed plugins spanning 13 working divisions, so you install only the ones you need.
What Matt's using now
The skills from this pack in my own daily setup, synced to the versions I actually run (last checked November 2026). If you only install a few, start here.
| Skill | Plugin | What it does for me |
|---|---|---|
| next | m2ai-workflow-content | Ends every session with a handoff note (where we are, what we decided, next step) so the next session or another tool resumes cold. |
| goal-maker | m2ai-workflow-content | Turns a half-formed idea into a runnable goal card with success criteria, budget, human gates, and how to run it. |
| decompose-goal | m2ai-workflow-content | Breaks a clean goal into atomic, dispatchable subtasks for subagents, workflows, or a checklist. |
| gh-review | m2ai-workflow-content | Reviews a GitHub repo against my current project: what it is, whether it fits, how I would integrate it. |
| swarm-discovery | m2ai-build-tooling | Maps an unfamiliar repo with a verified agent swarm before I plan anything against it. |
| get-api-docs | m2ai-build-tooling | Pulls current API docs before writing code against an SDK, so model names and parameters come from docs, not memory. |
| skill-maintenance | m2ai-build-tooling | Audits my skills for description quality, triggers, and structure. |
| model-router | m2ai-model-economics | Picks the model tier for a task and dispatches it, with logging so routing improves from outcomes. |
| model-audit | m2ai-model-economics | Scans a codebase for every model it calls and flags tier, task-fit, and cost changes. |
Install the three plugins that cover these:
/plugin marketplace add m2ai-portfolio/m2ai-skills-pack
/plugin install m2ai-workflow-content@m2ai-skills-pack
/plugin install m2ai-build-tooling@m2ai-skills-pack
/plugin install m2ai-model-economics@m2ai-skills-pack
Install
The pack ships as 7 themed plugins from this one repo. Install only the ones you want — each is independent.
| Plugin | Skills | What it covers |
|---|---|---|
m2ai-agent-architecture | 27 | Design, scaffold, and structure single- and multi-agent systems. Context hygiene, memory architecture, and world-model design. |
m2ai-safety-verification | 28 | Guardrails, permissions, kill switches, and judgment gates before an agent acts. Hostile-reviewer passes and trust scoring before anything ships. |
m2ai-agent-ops | 17 | Health checks, diagnostics, post-mortems, and run observability. |
m2ai-model-economics | 28 | Routing, migration, cross-model review, and provider monitoring. Token spend, boot tax, and inference-cost exposure. Lock-in scoring, licensing, and exit-cost analysis. |
m2ai-build-tooling | 38 | Spec-first build loops, code comprehension, and migration safety. Build, audit, convert, and maintain skills, tools, and MCP servers. |
m2ai-strategy-analysis | 27 | Briefings, stress-tests, market signals, and workflow-investment scoring. |
m2ai-workflow-content | 20 | Sessions, handoffs, recall, calendar, and remote access. Image and video prompt engineering and shorts pipelines. |
Claude Code
In Claude Code, add the marketplace once, then install the plugins you want:
/plugin marketplace add m2ai-portfolio/m2ai-skills-pack
/plugin install m2ai-agent-ops@m2ai-skills-pack
Or from your shell, without opening Claude Code first:
claude plugin marketplace add m2ai-portfolio/m2ai-skills-pack
claude plugin install m2ai-agent-ops@m2ai-skills-pack
To pre-declare the pack for a team (no install command at all), add to .claude/settings.json:
{
"extraKnownMarketplaces": {
"m2ai-skills-pack": { "source": { "source": "github", "repo": "m2ai-portfolio/m2ai-skills-pack" } }
},
"enabledPlugins": { "m2ai-agent-ops@m2ai-skills-pack": true }
}
Restart Claude Code. Skills auto-trigger on their descriptions, or invoke explicitly (e.g. /model-router, /diagnose, /aar).
There is no "Add to Claude Code" button, and no claude:// install link — Claude Code has no plugin-install deeplink. The commands above are the shortest path that exists.
Cursor
The same 7 plugins are also valid Cursor plugins (.cursor-plugin/), and Cursor auto-discovers skills/ with no extra declaration. Cursor's one-click install comes from being listed on the Cursor Marketplace, which is a manual review process.
Status: the Cursor install path is UNVERIFIED. The manifests are validated against Cursor's documented schema and against the shape of
cursor/plugins' own livemarketplace.json, but no Cursor install has been performed and no marketplace listing exists yet. Treat Cursor support as untested until this note says otherwise.
Upgrading from the old single-plugin install
If you installed m2ai-skills-pack@m2ai-skills-pack, it will break. That plugin no longer exists: the 183-skill monolith was split into the 7 plugins above.
There is no automatic migration. Claude Code's renames field maps one old plugin name to one new name, and cannot express one plugin becoming seven — so nothing in the format can do this for you. Uninstall the old plugin and install the themed ones you want:
/plugin uninstall m2ai-skills-pack@m2ai-skills-pack
/plugin install m2ai-agent-ops@m2ai-skills-pack
Your skills are all still here, under the same names, and none of their content changed — only which plugin ships them.
Starter Kit
New to Claude Code agents? templates/starter-kit/ contains a ready CLAUDE.md + agent.yaml pair that shows how the two files connect — useful for anyone setting up their first agent.
| File | What it is |
|---|---|
CLAUDE.md.template | A complete CLAUDE.md skeleton — hard rules, learned-mistakes log, adoption gate, and agent team rules — showing how behavioral constraints are written as enforced code, not documentation |
agent.yaml.example | A single agent.yaml identity card — model, description, tools, and routing fields — showing how it pairs with CLAUDE.md to form one complete agent |
Copy CLAUDE.md.template to ~/.claude/CLAUDE.md and agent.yaml.example to agents/<your-agent-name>/agent.yaml, fill in your details, then install the skills above. The rules/ subfolder contains starter rule files for shell safety and architecture discipline that CLAUDE.md.template already points to.
What's inside
| Division | Skills | Focus |
|---|---|---|
| 🏗️ Agent Architecture & Design | 19 | Design, scaffold, and structure single- and multi-agent systems. |
| 🛡️ Safety, Governance & Control | 18 | Guardrails, permissions, kill switches, and judgment gates before an agent acts. |
| 🩺 Operations, Reliability & Observability | 17 | Health checks, diagnostics, post-mortems, and run observability. |
| 🔀 Model & Inference Strategy | 12 | Routing, migration, cross-model review, and provider monitoring. |
| 💰 Cost & Token Economics | 6 | Token spend, boot tax, and inference-cost exposure. |
| 🔓 Portability & Vendor Independence | 10 | Lock-in scoring, licensing, and exit-cost analysis. |
| 🧠 Context & Memory | 8 | Context hygiene, memory architecture, and world-model design. |
| 🏭 Build Pipelines & Code Comprehension | 16 | Spec-first build loops, code comprehension, and migration safety. |
| 🧰 Skills, Tools & MCP | 22 | Build, audit, convert, and maintain skills, tools, and MCP servers. |
| 📊 Strategy & Business Analysis | 27 | Briefings, stress-tests, market signals, and workflow-investment scoring. |
| ✅ Deliverable Quality & Verification | 10 | Hostile-reviewer passes and trust scoring before anything ships. |
| 🗂️ Productivity & Workflow | 16 | Sessions, handoffs, recall, calendar, and remote access. |
| 🎬 Content & Media | 4 | Image and video prompt engineering and shorts pipelines. |
| 185 |
The catalog
🏗️ Agent Architecture & Design
Design, scaffold, and structure single- and multi-agent systems.
| Skill | What it does | When to use |
|---|---|---|
| agent-architecture-audit | Evaluate an agent codebase against 12 infrastructure primitives (permission model, token budget, crash recovery, tool… | Auditing agent architecture, reviewing agent readiness, or planning what infrastructure to build next |
| agent-sdk-layer-stack | Build a personal multi-agent AI command center using Claude Code as the immutable foundation with independently swapp… | Designing a personal AI stack, replacing an agent framework, building a multi-agent system, or asking "how do I build my own ag… |
| agents-md-generator | Generate or lint an AGENTS.md file from repository analysis. | The user says "generate agents.md", "create agents.md", "lint agents.md", "agents file", "agent instructions", or wants to crea… |
| anticipation-gap-audit | Scores any agent definition or skill manifest against the four consumer-AI breakthrough problems (context, reliabilit… | Designing or restructuring an agent system. |
| bitter-lesson-scorecard | Score an agent system design against the Bitter Lesson principle — how much "how" is encoded vs "what", how much bets… | Designing new agent systems, reviewing agent architecture, or deciding what to simplify |
| build-spec-generator | Takes a workflow that scored BUILD from workflow-fit-scorer and emits a deployment-ready build spec — trigger, I/O co… | Designing or restructuring an agent system. |
| classify-agent | Classify a problem into one of four agent architectures: coding harness, dark factory, auto research, or orchestratio… | Starting a new agent project, evaluating a build approach, or when someone says "what kind of agent should I build?", "classify… |
| claudemd-router | Restructure a bloated CLAUDE.md into a lean router that delegates to compartmentalized .claude/rules/ files. | CLAUDE.md exceeds ~80 lines, when adding a new domain of instructions, or when the user says "clean up my CLAUDE.md", "my CLAUD… |
| compound-skill-orchestrator | Build orchestrator skills that chain multiple existing skills in sequence using context:fork, passing output between… | The user says "chain skills", "combine skills", "orchestrator skill", "compound skill", "run skills in sequence", "skill pipeli… |
| decomposition-scorer | Score whether a set of coding tasks are properly decomposed on boundaries of isolation. | Launching parallel agent workers, when planning task breakdown for yce-harness or similar, or when someone says "will these tas… |
| delegation-spec | Generate a pre-flight delegation spec for any agent task -- context packages for memory-weak tools, review gates for… | Handing off a task to an agent, preparing a workflow for autonomous execution, or writing deployment specs for client agent sys… |
| dispatch-handoff-brief | Generate structured delegation briefs for Anthropic Dispatch or any autonomous agent handoff with objective, success… | Designing or restructuring an agent system. |
| dynamic-workflow-orchestration | Invoke Claude Code's dynamic workflow orchestration to coordinate teams of 10–50+ agents for large-scale, comprehensi… | Trigger phrases: "build a workflow", "dynamic workflow", "/workflows", "spin up agents for", "comprehensive audit across", "wor… |
| heartbeat-generator | Takes a user's described operating rhythms (from structured elicitation, manual input, or existing CLAUDE.md) and gen… | The user says "generate heartbeat", "build my schedule", "create cron from rhythms", "heartbeat.md", "schedule my agent", or wh… |
| hive-mind-shared-context | Design a shared context store for a multi-agent team -- write policies, decay rules, pinning, and consolidation sched… | Building a multi-agent system where agents need to share memory, designing memory decay and consolidation for an agent team, or… |
| make-anticipatory | Refactor any reactive slash-command skill or repeating manual workflow into an anticipatory one that fires automatica… | Saying "make this anticipatory", "convert this to a hook", "make it fire automatically", "this skill never fires because I forg… |
| prompt-goodies | An escalating 8-rung ladder of copy-paste delegation-prompt patterns, extracted from prompts a frontier model wrote for itself when orchestrating subagents. Serves rungs at your level, or upgrades a delegation prompt you paste in. | Writing or improving any prompt that hands work to an agent; an agent ignored instructions, touched the wrong files, or claimed success without proof; "prompt goodies", "make my subagent prompt better" |
| scheduled-agent-harness | Define and document a complete scheduled-agent contract for any recurring automated task. | Designing or restructuring an agent system. |
| sensemaking-concentrator | Audit a multi-agent system for distributed sensemaking anti-patterns and recommend where to concentrate interpretatio… | Designing or restructuring an agent system. |
🛡️ Safety, Governance & Control
Guardrails, permissions, kill switches, and judgment gates before an agent acts.
| Skill | What it does | When to use |
|---|---|---|
| action-class-policy | Interview the user about their agent usage patterns and risk tolerance, then generate a tiered action-class policy do… | Someone says "/action-class-policy", "define action classes", "agent permissions policy", "what actions can my agent take unsup… |
| agent-authority-map | Audit and visualize what an agent is actually allowed to do — parses Claude Code settings, hooks, MCP server configs… | The user says "agent authority map", "what can my agent do", "permission audit", "authority gaps", or wants to inventory autono… |
| agent-blast-radius | Inventory everything an autonomous agent has actually touched over a time window — files modified, commits authored… | The user says "blast radius", "what has my agent done", "agent footprint", "audit agent actions", or wants visibility into the… |
| classify-loop | Classify an agent workflow by the loop it operates in (code-generation, operational, supervisory, or fully-autonomous… | Designing a new agent workflow, auditing an existing agent deployment, or when someone asks "what loop is my agent in?", "how a… |
| claudex-plan | Enforced multi-round planning loop with a stop hook that blocks code execution until the plan survives adversarial re… | A task is high-stakes enough that a wrong plan is expensive to undo — the stop hook is the "bouncer" that prevents jumping stra… |
| control-map | Walk any agent workflow through a 7-row control map (runtime, governed data, identity/principal, action authorization… | Designing a new agent workflow, auditing an existing one, or answering "what controls does this agent have?" before shipping |
| describability-gate | Eight-field readiness gate that blocks any automation project until inputs, outputs, decision rules, exceptions, owne… | Kicking off any automation build, as a pre-dispatch check, or when reviewing automation specs |
| destructive-op-guard | Design or audit a PreToolUse hook that intercepts high-blast-radius shell operations (DROP TABLE, rm -rf, git push --… | Adding a safety hook to a Claude Code project, reviewing an existing hook's coverage, or after an incident where an autonomous… |
| info-judgment-boundary-auditor | Audit an agent system, workflow, or AI implementation to map where information retrieval ends and judgment calls begin. | Reviewing agent designs, deploying AI into decision workflows, or diagnosing "it was working fine then decisions got worse" fai… |
| judgment-gate | A judgment classifier for planned tool calls or agent actions. | Scores reversibility, cost boundary, blast radius, and user observability, then routes to ACT / ASK / QUEUE / ABORT with reasoning |
| judgment-layer | Add a judgment classifier to any agent or Claude Code skill to decide when to act autonomously vs. ask for confirmation. | Asking "when should my agent ask?", "judgment layer", "add a confirmation gate", "my agent acts when it shouldn't", "reversibil… |
| kill-switch-audit | Verify that a named agent in production has a real, one-step kill path at each of 5 layers (runtime cancel, credentia… | Use before shipping a new autonomous agent, during a security review, or any time you need to answer "can we actually stop this… |
| management-function-audit | Takes an org change description and classifies which management functions (routing, sensemaking, accountability) were… | Predicts failure modes and recommends mitigations based on historical precedents |
| plan-stop-hook | Installs a "bouncer" hook in Claude Code that blocks all file edits until an explicit planning gate is passed. | A task is non-trivial, prior attempts drifted, or the cost of a wrong plan is high |
| policy-gen | Generate the policy agent YAML security policies for agent sandboxing. | The user says "the policy agent policy", "agent sandbox", "generate security policy", "sandbox config", "the policy agent yaml"… |
| spec-driven-dev-enforcer | Enforce a spec-first development workflow that gates code generation behind explicit specification approval. | The user says "spec first", "spec driven", "write the spec before coding", "requirements first", "design doc first", "no code u… |
| tie-breaker | Break a tie when two agents or vendors disagree on how to clear a blocker: restate it in plain language, present both sides, synth… | Someone says "tie-breaker", "break the tie", "two models recommend different things", or two agents have given conflicting rec… |
| vendor-pressure-test | Run a vendor pitch, RFP response, or internal proposal through the 7-row agent control-layer lens to identify which r… | Evaluating a vendor for an agent deployment, responding to an RFP, or stress-testing your own internal proposal before presenti… |
🩺 Operations, Reliability & Observability
Health checks, diagnostics, post-mortems, and run observability.
| Skill | What it does | When to use |
|---|---|---|
| aar | Run an After-Action Review on an agent dispatch run, build execution, or multi-step orchestrated operation. | The user says "aar", "after action review", "let's review what happened", "review this run", or after any dispatch that produce… |
| action-outcome-postmortem | Structured post-mortem for the failure mode where an agent executed the correct mechanical action but produced the wr… | An agent did exactly what it was told but the result was wrong, or the user says "/action-outcome-postmortem", "right action wr… |
| agent-cold-start-diagnostic | Evaluate an existing agent setup (CLAUDE.md, SOUL.md, memory files, cron config, agent.yaml) and score it on operatio… | The user says "cold start diagnostic", "check agent config", "agent health score", "is my agent configured properly", "audit ag… |
| agent-correction-capture | Document and implement the pattern for capturing user corrections to agent output as structured events (user_correcti… | Building any agent-facing UI, CLI, or chat interface where you want corrections to flow into an evaluation or feedback pipeline… |
| agent-doctor | Run a comprehensive health check on all connected services, API credentials, MCP servers, tools, and external depende… | Validates that everything the agent needs is reachable and authenticated before the user hits a failure |
| agent-environment-optimizer | Audit an agent's execution environment for cold-start patterns, missing warm caches, stale dependencies, and session… | The user says "optimize agent environment", "cold start audit", "agent environment check", "why is my agent slow to start", "wa… |
| agent-run-schema | Reference skill containing the canonical 10-event agent analytics schema. | Defines event shapes (agent_run_started, task_completed, user_correction_submitted, approval_granted, approval_denied, tool_cal… |
| agent-standup | Run a structured standup across a multi-agent team. | Asking for a "standup", "team status check", "/standup", "war room update", "what are all my agents doing?", or "agent status r… |
| completion-vs-acceptance | Analyze the gap between an agent's self-reported completion rate and the actual acceptance rate (tasks the user or do… | The user says "completion rate", "acceptance rate", "how trusted is my agent", "agent done vs actually accepted", "completion-v… |
| diagnose | Strict 4-gate diagnostic protocol for bugs, failures, and broken pipelines. | The user reports a bug, something is broken, a test is failing, a service is down, or says "diagnose X", "what's wrong with X"… |
| failure-asymmetry | Compare a skill's behavior under human invocation vs simulated agent invocation, highlighting divergences in output f… | Diagnosing, reviewing, or monitoring agent runs. |
| failure-postmortem | Guide a structured AI system failure post-mortem using 6 named failure patterns, producing a publishable incident rep… | Diagnosing, reviewing, or monitoring agent runs. |
| hallucination-audit-trail | Cross-checks an agent's self-reported processing log against filesystem evidence (file modification timestamps, exist… | Diagnosing, reviewing, or monitoring agent runs. |
| open-loop-audit | Audit open loops and classify each as real delegation candidate vs simulated work using Nate's "land or leave" framework | Diagnosing, reviewing, or monitoring agent runs. |
| session-work-log | Produces a structured session-end artifact capturing what was attempted, what changed, what blocked progress, and wha… | Machine-readable by design -- survives model swaps and cross-agent handoffs |
| simulated-work-detector | Recurring audit that reviews agent fleet output and flags simulated work that generated artifacts but didn't close an… | Diagnosing, reviewing, or monitoring agent runs. |
| trace-prompt | Reconstruct a sanitized diagnostic trace of the current prompt exchange only and write it to a markdown file for shar… | The user says "trace this", "trace-prompt", something just went wrong, or needs a concrete bug report instead of a vague compl… |
🔀 Model & Inference Strategy
Routing, migration, cross-model review, and provider monitoring.
| Skill | What it does | When to use |
|---|---|---|
| adversarial-dual-model-workflow | Set up and run a two-AI coding workflow where a primary assistant handles planning and code generation while an adver… | The user wants complementary AI tool pairing, a structured code review gate, an adversarial planning loop, or a background audi… |
| api-breaking-change-scanner | Scans a codebase for deprecated API parameters and patterns that break silently or return 400 errors when migrating t… | Choosing, migrating, or reviewing models. |
| compensating-complexity-auditor | Audit system prompts and agent pipelines for compensating complexity — scaffolding, procedural hacks, and duct tape b… | Upgrading models, reviewing system prompts, or preparing for model migration |
| compute-availability-tracker | Poll the status pages of major AI providers (Anthropic, OpenAI, Google, OpenRouter) and produce a current availabilit… | A model call is failing, before routing important work, or on a schedule to detect degradation early |
| cross-model-peer-review | Implements a cross-model peer review loop where Model A grades Model B's output on a structured rubric, catching erro… | Choosing, migrating, or reviewing models. |
| gemini-rescue | Get a second opinion from Google Gemini on the current diff, plan, or problem. | Claude is stuck, when an implementation needs adversarial review by a different vendor, or when a 1M-context audit is needed th… |
| model-audit | Benchmark and compare model assignments across a multi-stage pipeline; produce a model-by-stage audit report with cos… | Choosing, migrating, or reviewing models. |
| model-migration-preflight | Audit your prompts, system instructions, and Claude Code configuration before switching model versions. | Any model version upgrade that affects production workflows |
| model-release-deep-dive | Produce a structured capability-delta brief for any AI model release -- what actually changed beyond benchmark scores… | A new model release or when benchmarks alone don't explain real-world behavior changes |
| model-router | Classify a task, pick the model tier (Fable/Opus/Sonnet/Haiku/local), and dispatch it, with route and outcome logging. | Choosing, migrating, or reviewing models. |
| prompt-rewriter | Rewrite system prompts to strip compensating complexity and produce clean outcome-based prompts. | Modernizing prompts after a model upgrade, simplifying over-engineered system prompts, or converting procedural instructions to… |
| provider-availability-scout | Scheduled task that probes model provider availability, pricing changes, restriction announcements, and API deprecati… | Diffs against last run and optionally updates a routing defaults file so your model routing layer stays current without manual… |
💰 Cost & Token Economics
Token spend, boot tax, and inference-cost exposure.
| Skill | What it does | When to use |
|---|---|---|
| agent-cost-model | Model token costs and optimization opportunities for any agent workflow, producing per-task costs, monthly burn, and… | When token spend or session cost needs auditing. |
| ai-cost-exposure-audit | Walk through a three-prompt sequential audit of your organization's AI inference cost-structure exposure -- inventory… | You want to map your exposure to the AI inference cost curve before it breaks your budget |
| boot-tax-monitor | Monitor and alert when Claude Code session startup overhead (plugins, skills, MCP servers, CLAUDE.md chain) exceeds a… | Prevents the silent context bloat that eats your working memory before you type a single prompt |
| context-cost-line-item-analyzer | Decompose an agent's token spend into context-as-input vs output-as-output, identify the specific context drivers (st… | The user says "context cost breakdown", "why is context eating my budget", "context line items", "decompose my token spend", "w… |
| pre-turn-budget-guardian | Enforce a token budget ceiling on Claude Code sessions by checking projected usage BEFORE each turn and halting with… | Prevents runaway loops and silent token burn |
| token-burn-auditor | Audit the live Claude Code environment for token waste -- measures per-session overhead, flags system prompt bloat, c… | Real-time linter for AI workflows |
🔓 Portability & Vendor Independence
Lock-in scoring, licensing, and exit-cost analysis.
| Skill | What it does | When to use |
|---|---|---|
| agent-memory-architecture-matrix | Interactive decision matrix that walks a builder through choosing between provider-hosted memory (fast, locked) and s… | The user is designing an agent's memory layer, comparing memory backends, deciding between Claude memory, Mem0, Zep, SQLite, or… |
| cross-runtime-skill-converter | Convert a Claude Code skill to Codex format or vice versa. | : "convert my skill to Codex", "port this skill to Claude Code", "make this skill work in both runtimes", "cross-runtime skill"… |
| fair-agent-license-rubric | Score an agent license or SaaS contract term sheet against a 9-trait rubric distinguishing fair agent licensing (tran… | Assessing lock-in or planning a vendor exit. |
| inference-layer-ownership-auditor | Map your AI stack to identify who controls each inference layer -- model provider, hosting, hardware -- and score sub… | A strategic AI vendor decision or during portability planning |
| inference-vendor-lock-in-scorer | Scan agent and skill manifests for AI vendor lock-in exposure -- hardcoded model names, non-portable API surfaces, ab… | A vendor migration or portability planning sprint |
| license-audit | Takes a team's current AI tool inventory and work types, then outputs a one-page misfit memo identifying duplicative… | Interactive advisory skill for AI tool stack rationalization |
| mcp-portability-auditor | Scan an agent stack (Claude Code plugins, Conway .cnw.zip files, Gemini extensions, ChatGPT GPTs) and flag which capa… | The user asks to audit their agent extensions, score MCP portability, check for vendor-proprietary extension formats, or plan a… |
| platform-dependency-mapper | Audit an org's or individual's AI stack and produce an exit-cost estimate across four axes — data, integrations, beha… | The user asks to audit AI lock-in, map vendor dependencies, estimate switching costs, or assess platform risk on their AI stack |
| poly-skill | Convert any Claude Code skill to work in OpenAI Codex (or vice versa) by applying a structural adapter that handles n… | Assessing lock-in or planning a vendor exit. |
| renewal-interrogation | Generate a vendor-specific SaaS renewal playbook for contracts where agent workloads are in scope. | The user says "renewal-interrogation", "renewal playbook", "SaaS renewal prep", "contract negotiation for agents", "help me pre… |
🧠 Context & Memory
Context hygiene, memory architecture, and world-model design.
| Skill | What it does | When to use |
|---|---|---|
| claude-code-memory-architect | Interviews the user to design and build a personalized Claude Code memory system — choosing which building blocks (de… | A user wants to build, redesign, or consolidate their Claude Code memory layer |
| claude-memory-architect | Interview-driven skill that designs a custom Claude Code memory system tailored to how the user actually works. | Clones open-source memory repos, audits their patterns, then builds a personal memory spec using building blocks — decay, promo… |
| context-fork-guide | Use when creating or modifying skills that do heavy research, search, or multi-tool work. | Creating or modifying skills that do heavy research, search, or multi-tool work |
| context-hygiene | Reference for Claude Code context management. | "context is getting large", "session feels slow", "Claude forgot what we were doing", "how do I manage context" |
| memory-fingerprint-architect | Design a personalized Claude Code memory system by cloning existing open-source memory frameworks, auditing them with… | Saying "design my memory system", "build my memory", "personalize my CLAUDE.md memory", "memory architect", or "memory fingerpr… |
| memory-system-assembler | Build a personalized Claude Code memory system by auditing open-source memory repos, cherry-picking patterns that mat… | Setting up memory for the first time, redesigning an existing memory system, or evaluating new memory patterns from the OSS eco… |
| world-model-principles-scorecard | Score an agent system or knowledge base against five principles that determine whether a world model compounds value… | Complementary to bitter-lesson-scorecard (architecture simplicity) — this measures knowledge/decision quality |
| world-model-readiness-diagnostic | Interactive diagnostic that maps an organization to a world model paradigm (vector DB, structured ontology, or signal… | Assessing organizational AI readiness, choosing a knowledge architecture, or starting a consulting engagement |
🏭 Build Pipelines & Code Comprehension
Spec-first build loops, code comprehension, and migration safety.
| Skill | What it does | When to use |
|---|---|---|
| agent-readiness-audit | Audit a codebase for agent-readiness across 8 pillars (style/validation, build systems, testing, docs, dev environmen… | The user says "audit readiness", "agent readiness", "codebase audit", "is this repo ready for agents", "readiness score", or wa… |
| claude-architect-audit | Audit your Claude Code setup against Anthropic's 5 Certified Architect domains. | You want to level up your Claude Code configuration, prepare for the architect certification, or diagnose why Claude isn't perf… |
| claude-design-to-code | Rapidly prototype UI wireframes and high-fidelity designs with Claude Design, then export to Claude Code with a singl… | Saying "design to code", "wireframe to code", "prototype in Claude Design", "export Claude Design", or "Claude Design handoff" |
| code-review-memory | Accumulates repo-specific code review lessons and surfaces them before the next review of the same file or module. | Learns from recurring issues -- migration bugs, error-handling patterns, fixture quirks, security gaps -- so the same mistake i… |
| comprehension-gate | Pre-merge review that checks a PR diff not for syntax or test coverage, but for human understanding of implications -… | The user says "comprehension gate", "comprehension check", "understanding review", "implication review", "pre-merge comprehensi… |
| comprehension-interview-loop | Conduct a Socratic pushback interview about something the user has built. | The user says "interview me", "comprehension interview", "Socratic review", "pushback interview", "do I actually understand thi… |
| context-layer-generator | Generate three structured context artifacts for any module or service -- a structural manifest (what it does, depende… | The user says "generate context layer", "context layer", "module context", "behavioral contracts", "decision log", "document th… |
| dark-code-audit | Scan a codebase to map where human comprehension has gone missing -- identifying modules with no decision logs, no be… | Onboarding to a new repo, auditing AI-generated code risk, assessing technical debt from comprehension gaps, or as a periodic h… |
| explanation-artifact-generator | Walk through a 4-question comprehension template for any project artifact (repo, feature, tool) and produce a structu… | The user says "explanation artifact", "explain this project", "comprehension template", "4-question template", "what did I lear… |
| image-to-implementation | Generate a visual reference image for a UI or design concept, then produce a structured implementation prompt that ha… | Closes the "AI invents bad taste" failure mode by anchoring implementation to a concrete visual rather than a text description… |
| mismatch-check | Diagnose whether your current agent tooling matches your actual problem. | Evaluating an existing agent setup, troubleshooting why an agent isn't delivering, or when someone says "why isn't this working… |
| self-healing-claudex | Self-healing build pipeline with Planner/Builder + Codex adversarial review. | The user says "self-healing claudex", "claudex build", "PBJ with codex review", "two-vendor build loop", or wants Codex (not ju… |
| self-healing-pipeline | Self-healing build pipeline with Planner/Builder/Judge loop. | The user says "self-healing pipeline", "healing build", "auto-fix pipeline", "build with retries", or wants a resilient impleme… |
| spec-gap-detector | Stress-test any agent prompt or specification for ambiguity, missing constraints, and edge cases that would cause ran… | During spec-first builds or code review. |
| swarm-discovery | Map an unfamiliar repo with a verified agent swarm: Haiku maps each module, Sonnet skeptics vote on claims, Opus writes the architecture map. | First contact with a repo or service tree, before a refactor, migration or ETL design (needs the Workflow tool) |
| validated-data-migration | Migrate a shoebox of inconsistent files (CSV, Excel, JSON, PDFs, VCF) into a clean SQLite database with rejection log… | Onboarding a client with legacy data or cleaning up a data dump before importing into a new system |
🧰 Skills, Tools & MCP
Build, audit, convert, and maintain skills, tools, and MCP servers.
| Skill | What it does | When to use |
|---|---|---|
| agent-commerce-mcp-spec | Take a business's existing storefront, catalog, or checkout flow and emit a complete MCP server spec that makes it ag… | Asked "make this storefront agent-callable", "generate MCP spec for my shop", "agent commerce spec", "MCP server for checkout"… |
| agent-commerce-scaffold | Scaffold a new agentic-commerce product with all 8 commercial-responsibility domains pre-stubbed -- Identity, Authori… | The user says "scaffold an agent product", "agent commerce scaffold", "init agentic product", "commerce agent init", "agent com… |
| agent-infra-scorer | Score any enterprise tool (Jira, Salesforce, Workday, Google Calendar, etc.) against a 5-test structural diagnostic t… | The user says "score my stack", "is Jira agent-ready", "agent infrastructure test", "which tools should I MCP-wrap", or wants t… |
| agentify-docs | Converts existing human-oriented documentation (READMEs, runbooks, architecture docs, onboarding guides) into agent-r… | Asked to "write AGENTS.md", "make this doc agent-readable", "add agent instructions", "convert our runbook for agents", or "/ag… |
| automation-platform-decision-guide | Decision-support skill for choosing between agent automation platforms. | Given a use case and org context, recommends between scheduling platforms (n8n, Make, Workspace Agents), custom agent framework… |
| book-to-skill | Create Claude Code skills from book/long-form prose. Reads a how-to chapter, extracts the teachable SOP, then converts it into a skill via skill-creator. Prose sibling of video-to-skill. | The user says "book to skill", "chapter to skill", "extract skill from this chapter", "turn this how-to into a skill", or "SOP from this writing" |
| callable-audit | Audit whether a business is callable by an AI agent end-to-end. | Asked "can an agent buy from this business", "is this service agent-callable", "callable audit", or "agent-commerce readiness" |
| demotion-audit | Three-phase audit to determine whether a software tool or artifact should be demoted down the production-class ladder… | "demotion audit", "should I demote this tool", "is this tool still worth supporting", "downgrade this tool", "retire this tool" |
| eval-agent | Score any agent tool or platform against 3 structural questions (persistent memory, inspectable artifacts, compoundin… | Evaluating agent tools, comparing platforms, or deciding whether to trust a tool with a specific workflow |
| failure-mode-tool-router | Given any task description, classify it into the right AI tool category and explain why common alternatives are wrong… | Prevents wasting time building agents for tasks that should use a different tool entirely |
| get-api-docs | Fetch authoritative API documentation via the Context Hub (/chub) to verify model names, parameters, and endpoints be… | Building, auditing, or converting skills and tools. |
| inbox-load-audit | Scans installed skills, plugins, MCP servers, hooks, and scheduled tasks to produce an inbox load score — how many in… | Building, auditing, or converting skills and tools. |
| mcp-spec-generator | Takes any SaaS product or internal system and outputs a complete MCP server specification — tool list, input/output s… | The user says "spec an MCP server for X", "write an MCP spec", "how would I wrap X as MCP", or wants a buildable blueprint befo… |
| optimize-description | Rewrite a skill's description field for maximum agent discoverability -- clear trigger conditions, explicit input/out… | Building, auditing, or converting skills and tools. |
| prototype-classifier | Classify any software tool, script, or AI artifact onto a 4-rung production-class ladder (Personal Tool → Team Beta →… | Never rounds up on aspiration — only classifies at the highest rung the artifact fully qualifies for |
| skill-audit | Audit your Claude Code skills directory, CLAUDE.md, and recent session patterns to surface candidates for new skills… | Building, auditing, or converting skills and tools. |
| skill-maintenance | Audit Claude Code skills for content quality against Anthropic best practices. | A scheduled maintenance cycle runs, when the user asks to check skill quality, or when reviewing skills before publishing |
| tool-audit | Audit a tool already in your stack against a semantic-depth rubric and emit a clear decision — Extend, Wrap, Replace… | Evaluating whether a current tool is worth keeping, improving, or replacing; when a tool is causing agent failures; or the user… |
| tool-build-guide-generator | Generate a hands-on, machine-ready build guide for any tool or library — takes a tool name and produces a runnable, a… | Not a hello-world demo; a real build that develops genuine fluency |
| tool-fluency-builder | Turn any AI tool or framework from "heard of it" to "can build with it" by generating a machine-ready, hands-on build… | Goes beyond hello-world to produce a real, runnable project with annotated steps, pitfalls, and judgment calls |
| tool-migration-brief | Converts enterprise tool audit results (which tools fail agent-infrastructure tests) into a leadership-ready migratio… | The user says "write a migration brief", "make this exec-ready", "brief for leadership on tool migration", or has scored their… |
| video-to-skill | Create Claude Code skills from screen recordings of manual processes. | The user says "record to skill", "video to skill", "screen record skill", "extract skill from video", "SOP from recording", or… |
📊 Strategy & Business Analysis
Briefings, stress-tests, market signals, and workflow-investment scoring.
| Skill | What it does | When to use |
|---|---|---|
| agent-system-touch-map | Map every workflow to the SaaS systems it touches, the operations performed on each system, and the vendor billing me… | The user says "agent-system-touch-map", "map agent costs", "SaaS meter map", "what will this agent cost to run", "agent licensi… |
| amdahl-ceiling-calculator | Map a workflow's time allocation between AI-accelerable and non-accelerable steps, calculate the theoretical maximum… | The user says "amdahl ceiling", "workflow bottleneck", "max speedup", "where's my ceiling", "why isn't AI faster", "calculate s… |
| arbitrage-audit | Map a business model to Nate's five exploitable-inefficiency categories (speed, reasoning, fragmentation, discipline… | The user wants an org-level diagnostic of where AI is about to collapse their margins or which parts of their business are most… |
| brand-memory-probe | Probe how a brand appears in AI agent recall across multiple LLM providers. | Asked "does my brand appear in AI search", "brand memory probe", "AI recall audit", "how does ChatGPT describe us", or "agent s… |
| buyer-questions | Pre-pitch preflight for enterprise AI sales. | Input a product description and buyer persona; outputs the two highest-probability deal-killing questions that buyer will ask… |
| career-gap-map | Calculate an individual's AI exposure score (percent of weekly tasks AI can already compress to near-zero) and produc… | The user wants to audit their own job, plan a career pivot, evaluate how exposed their role is to the next model release, or bu… |
| counterargument-stress-test | Takes any strategic argument, business case, or thesis and systematically generates and addresses the N strongest cou… | Analyzing a market, workflow, or business decision. |
| data-fabric-fit-assessor | Given a client's or team's existing data infrastructure (Microsoft 365, Google Workspace, Salesforce, custom RAG, etc… | Maps four major data fabrics to agent products optimized for each |
| decision-interview-builder | Conduct a structured multi-turn interview through one real past decision the user made — walking situation, decision… | The user says "interview me about a decision", "build my judgment artifact", "extract my reasoning", "decision interview", "sho… |
| executive-briefing | Takes a complex geopolitical, industry, or market event and produces a structured executive briefing with thesis, tra… | Analyzing a market, workflow, or business decision. |
| gap-trace | Run Nate's three-question industry diagnostic — what inefficiency is this built on, how fast can AI close it, what ne… | The user wants a fast (5-minute) diagnostic of any industry, product, or workflow without running the full arbitrage-audit |
| geopolitical-signal-enricher | Enriches a market or technology signal with geopolitical context -- affected regions, supply chain nodes, timeline es… | Analyzing a market, workflow, or business decision. |
| impl-arch-audit | Score an AI deployment against six implementation layers (workflow, data, authority, evaluation, audit trail, recovery). | Analyzing a market, workflow, or business decision. |
| launch-filter | Run any agent or AI product announcement through Nate's 5-question launch filter to get a structured verdict and go/w… | Evaluating a new agent platform, AI product release, or vendor announcement |
| middleware-trap-detector | Diagnose whether an agent deployment is falling into the "middleware trap" — wrapping legacy systems without redesign… | The user says "middleware trap", "deployment health", "is this deployment safe", "wrapping legacy", "automated bottleneck", or… |
| news-narrative-decomposer | Takes a month of AI/tech news and extracts the structural shifts underneath the headlines -- classifying each story b… | Fights "absorbed a lot of takes but couldn't name what changed |
| reasoning-extractor | Apply a four-question reasoning framework (situation → decision → risk → change) to any piece of work — from an indiv… | The user says "extract my reasoning", "reasoning trail", "document my thinking", "why I made this call", "annotate this decisio… |
| semantic-score | Score any AI product, vendor pitch, or announcement across 10 semantic-depth dimensions to distinguish "access-only"… | Evaluating AI tooling, reviewing a vendor pitch, auditing an integration, or the user says "/semantic-score", "semantic depth t… |
| silver-platter | Interview a business owner about their day-to-day tools, build a tailored data map, render a Pantry → Prep → Plate HT… | Audits existing Claude Code setups in the cwd before asking questions, so users who've started building don't get re-asked |
| strategic-timing-matrix | Map your biggest pending decisions (hiring, fundraising, product launches, home purchase, job change) against macro e… | The user says "timing matrix", "when should I", "decision timing", "should I wait", "accelerate or delay", "macro timing", "str… |
| stress-test-finder | Interview the user about their current workload, surface the task they've been avoiding because it felt too messy or… | Deciding what to hand off next or when stuck in indecision about AI delegation |
| thesis-cluster-detector | Ingests a batch of enterprise AI or technology news items (from RSS, Substack previews, or pasted headlines) and clus… | The user says "thesis cluster", "what's the underlying bet here", "cluster these stories", "detect the pattern", "what do these… |
| weekly-signal-diff | Tracks N companies across M categories, re-ranks using user context (projects, interests, priorities), and produces a… | Saves analysis back so signal compounds over time |
| workflow-decomposer | Turn a function, role, or process description into a list of discrete, scoreable workflows — the pre-step before any… | "decompose this role", "break this function into workflows", "workflow decomposer", or before scoring any AI automation opportu… |
| workflow-fit-scorer | Score a candidate workflow against 5 automation-fit properties and return a build / switch-tools / resolve-ambiguity… | Front-door diagnostic for any M2AI automation advisory engagement |
| workflow-investment-scorer | Score each workflow on six dimensions (frequency, mistake cost, judgment, model-improvement trajectory, market maturi… | "score this workflow", "investment motion", "should I automate or buy", "build buy hire wait" |
| workflow-tier-classifier | Paste a product description or pitch and get back a five-tier verdict (Decoration → Full Stack) with the missing impl… | Lightweight lead-magnet classifier and self-assessment tool |
✅ Deliverable Quality & Verification
Hostile-reviewer passes and trust scoring before anything ships.
| Skill | What it does | When to use |
|---|---|---|
| artifact-contract-validator | Validate a directory of claimed deliverables for file-type authenticity, formula presence in spreadsheets, embedded m… | Shipping any AI-generated deliverable set to a client or reviewer |
| institutional-taste-encoder | Formalize implicit quality judgments, style preferences, and "taste" into structured constraint specifications that a… | The user says "encode my taste", "formalize quality rules", "taste encoder", "write my style constraints", "capture my preferen… |
| knowledge-work-tests | Generate evaluation criteria, acceptance checks, and scoring rubrics for knowledge work outputs -- the missing "test… | Defining quality gates for agent outputs, building acceptance criteria for delegated work, or reviewing knowledge work delivera… |
| multi-artifact-work-package | Turn a messy business situation into a full deliverable set with an explicit artifact contract — Word, PowerPoint, Ex… | A client engagement, proposal, or agent task requires multiple coordinated output formats |
| pretty-but-wrong | Final hostile-reviewer pass on any AI-generated document, deck, or workbook before sharing. | Sharing any AI-generated deliverable with a decision-maker |
| promotion-packet-auditor | Adversarial reviewer that scores a promotion packet or performance self-review for real judgment signal versus AI-gen… | The user says "audit my promo packet", "signal check", "promotion review", "does this show real impact", "what will a reviewer… |
| recognizable-quality-test-generator | Generate an LLM-as-judge quality rubric from a workflow description and historical examples. | Building an agent that needs verifiable success criteria — converts "I know good when I see it" into a prompt-based pass/fail t… |
| source-packet | Builds a structured source inventory from a set of files or documents before any artifact is created. | Starting any AI-generated deliverable (deck, report, workbook) to establish what is known, what is estimated, and where sources… |
| trust-score | Score any AI-generated document, spreadsheet, or deck on a 0–100 trust scale across five verifiable dimensions. | Final review before sharing a deliverable. |
| workbook-doctor | Audits an existing or AI-generated Excel workbook for hidden risks before it is used in decisions. | Relying on any AI-generated spreadsheet for financial decisions, reporting, or client deliverables |
🗂️ Productivity & Workflow
Sessions, handoffs, recall, calendar, and remote access.
| Skill | What it does | When to use |
|---|---|---|
| afk | Overnight worker fired by a scheduled task. Drains a content-backlog inbox: expands items tagged afk: ready into dr… | A nightly scheduler fires it, or the user says "/afk", "run afk", "drain the inbox" |
| calendar-hygiene | Audit the next N days of a Google Calendar and surface hygiene issues before the user notices them -- back-to-back me… | Asking "calendar hygiene", "audit my calendar", "check my schedule", "what's wrong with my week?", "add buffers to my calendar"… |
| decompose-goal | Decompose a free-text goal into an ordered list of atomic, dispatchable subtasks for any downstream executor — subagents, Agent Teams, DAG engines (LangGraph, CrewAI, n8n), headless cron, or human checklists. Downstream of goal-maker. | A multi-step instruction needs atomization before handoff; "decompose this goal", "break this into subtasks", "flatten this for dispatch", "turn this into a runnable task list" |
| file-intel | Run the Gemini file processor on any folder — extracts content from PDF, PPTX, XLSX, DOCX, CSV, JSON, and any text fo… | Asked to "summarise this folder", "run file intel", "process these files", or a folder path is provided and summaries are needed |
| gh-review | Review 1-3 GitHub repos against the current project and generate an HTML report with plain-speak summaries, relevance… | Ever the user shares a GitHub repo link and wants to understand how it fits their project, asks 'what is this repo?', 'how woul… |
| goal-maker | Turn a fuzzy idea or triaged item into a well-formed, runnable GOAL card — crisp objective, observable done-state, and a chosen execution shape (one-shot/loop/cron/subagents/worktree) with owner/sink/kill. Upstream of decompose-goal. | A raw idea needs sharpening into an objective with success criteria and a way to run it; "make a goal", "turn this into a goal/loop", "how should I run this" |
| goal-os-optimize | Use /goal to run a self-directed optimization loop on your agentic OS — skills, CLAUDE.md, rules files, and projects. | A judge agent on a separate LLM validates each iteration |
| life-engine | Reference implementation of a proactive Life Engine -- a time-windowed briefing loop that checks email, calendar, pro… | This is a REFERENCE IMPLEMENTATION showing one user's full wiring, not a plug-and-play template |
| next | Write an end-of-session handoff (where we are, what we decided, next step) to ~/handoffs/<project>/ so any tool can resume. | Wrapping up a session or handing off work |
| quick | Quick-capture a spur-of-the-moment idea into a content-backlog inbox file so it is not lost. Tags captures for the AF… | The user says "/quick …", "quick capture", "jot this down", "add to the backlog", or fires off a half-formed idea |
| remote-channels | Set up and configure Claude Code remote access via Telegram and Discord channels. | Guides through bot creation, plugin installation, security lockdown, and always-on configuration |
| sparring-planner | Custom planning mode that challenges assumptions, asks deep multi-option questions, and explores multiple approaches… | The user says "plan", "let's think through", "help me decide", "sparring planner", "think this through with me", "let's figure… |
| structured-elicitation | A conversational skill that interviews the user across five layers of operational knowledge -- operating rhythms, rec… | The user says "elicitation", "interview me", "extract my workflows", "bootstrap agent persona", "build my SOUL.md", "help me do… |
| tldr | Save a summary of this conversation to the vault. | Key decisions, things to remember, next actions |
| transaction-history-builder | Aggregate completed projects, commits, PRs, shipped features, and explanation artifacts into a running transaction hi… | The user says "transaction history", "build my portfolio", "aggregate my work", "show my trajectory", "living resume", "proof o… |
| what-am-i-forgetting | Comprehensive agenda recall across all of your active projects, phases, gates, queued work, ideas, and cleanups. | Ever you ask "what's next", "what's on the agenda", "what am I working on", "what am I forgetting", "what's open", "what's in p… |
🎬 Content & Media
Image and video prompt engineering and shorts pipelines.
| Skill | What it does | When to use |
|---|---|---|
| banana-maker | Generate images using the Gemini image generation API with the Nano Banana Pro prompting methodology. | The user asks to generate, create, or make an image, picture, photo, illustration, or visual |
| seedance-prompt | Master AI Video Prompt Engineer for Seedance 2.0. | Converting a user concept into a high-quality, cinematic Seedance 2 |
| seedance-shot-prompt | Use when generating a Seedance 2 video prompt for a linear forward-motion shot — transitions, chase shots, establishi… | Generating a Seedance 2 video prompt for a linear forward-motion shot — transitions, chase shots, establishing shots, A→B narra… |
| viral-shorts-pipeline | Generate and publish viral TikTok/YouTube Shorts from a single theme prompt. | Asked to "make viral videos", "generate TikToks", "create shorts", "run the viral pipeline", or "generate and post videos" |
Configuration
Some skills expect environment variables for user-specific paths. If unset, each skill falls back gracefully and flags the missing source in its output.
| Variable | Used by | Purpose |
|---|---|---|
$PROJECTS_DIR | what-am-i-forgetting | Root of your code projects (e.g. ~/projects) |
$VAULT_DIR | what-am-i-forgetting, aar | Notes vault for daily notes and saved reports |
$MEMORY_INDEX | what-am-i-forgetting | Path to your memory-index file |
$MEMORY_DIR | what-am-i-forgetting | Directory containing memory files |
$ORCHESTRATOR_DB | aar | SQLite DB holding agent mission/task state |
$OPEN_ITEMS_JSON | aar | Queue file for action-item append |
GEMINI_API_KEY | file-intel, banana-maker | Gemini API access |
Known Rough Edges
A few skills contain environment-specific references. They still work, but you may want to adjust:
- gh-review — hardcoded LAN IP
<host>for the report-server URL. Swap for your own file server, or ignore (the HTML file is still generated locally). - banana-maker — one mention of a "Surface tablet." Harmless.
- context-fork-guide — example command points at
/home/user/.claude/skills/. Replace with your skills path if you run the example verbatim. - get-api-docs — depends on the
/chub(Context Hub) skill being installed separately. Without it, this skill is a no-op. - aar — assumes an orchestrated agent system with a SQLite task DB and A2A endpoints. Without those, Phase 1 forensics gets thin.
- what-am-i-forgetting — most useful if you maintain a memory-index file, per-project TODO/NEXT files, and daily notes. Without them, it falls back to folder inventory + cron listings.
- gemini-rescue — requires a Gemini MCP server exposing
gemini_askandgemini_runtools. Configure with your Google API key. - self-healing-claudex — requires the Codex CLI (
codex) installed and authenticated. Falls back to the all-Claude/self-healing-pipelineif Codex is unavailable. - tldr -- assumes an Obsidian vault at
~/vault/. Adjust the path in the skill if your vault lives elsewhere. - life-engine -- reference implementation, not plug-and-play. Shows how one user wired Gmail, Google Calendar, an idea-catcher DB, and changelog tracking into a 30-minute proactive loop. Read the Adaptation Guide at the bottom of the skill to build your own.
- silver-platter — uses Jinja2 for HTML rendering (
pip install jinja2). Examples use fictional business personas.
Version
0.5.1 -- added trace-prompt, a diagnostic skill that captures a sanitized markdown trace of the current exchange only (prompt, active model, tools run, errors verbatim) for support threads and bug reports (179 skills).
0.5.0 -- consolidated SkillForge into the pack (now 178 skills), sanitized for sharing, and reorganized the README into 13 divisions. Feedback welcome.