skill-maker

Run Microsoft SkillOpt experiments for agent skills, prompts, or instruction files with clean train/validation/held-out splits, adapter setup, bounded rollouts, artifact collection, and honest reporting. Use when the user wants to use SkillOpt, optimize a skill with SkillOpt, benchmark prompt variants, run multi-epoch skill optimization, check held-out quality, or diagnose SkillOpt runs.

Run pre-implementation AI consulting engagements—company research, discovery, maturity assessment, opportunity prioritization, and proposal/SOW/roadmap creation—and use only for advisory work, not implementation.

Guide annuity and life-insurance sales workflows—discovery, suitability, product selection, objections, closing, and servicing—when supporting licensed insurance sales conversations or deliverables.

Generate API reference documentation from source code—endpoints, schemas, auth, errors, and examples—when the task is to document an existing API rather than design or implement one.

Turns an approved build spec into a minimal technical implementation plan, including system components, human-in-the-loop boundaries, workflows, and delivery milestones. Use when an internal automation, concierge workflow, or lightweight product has already been chosen and now needs a concrete build plan.

Generate release notes and changelogs from git history with SemVer classification, grouped changes, and migration notes when documenting what changed between versions or refs.

Screens business ideas, offers, and product plans for employer IP, confidentiality, regulatory overreach, sensitive-data misuse, and unsafe claims, then proposes safer clean-room alternatives. Use when a user's edge may depend on employer information, regulated workflows, customer data, or legally sensitive positioning.

Build, validate, serialize, or parse CNCF CloudEvents v1.0 payloads and transports when the task explicitly involves CloudEvents envelopes, ce-* headers, or structured/binary event delivery.

Review existing code or diffs for bugs, security issues, performance problems, maintainability risks, and test gaps when the user wants evaluation, not new implementation.

Design safe, reversible database schema or data migrations with rollback, verification, and deployment guidance when the task changes an existing database structure or stored data.

Identifies differentiated money-making opportunities for skilled solo operators, ranks them across cash-now, asset-later, and hybrid paths, and produces validation plans plus tool/build specs. Use when a user wants business ideas, monetization strategies, micro-SaaS wedges, service-to-software paths, or AI-leveraged opportunities matched to their skills, assets, access, credibility, constraints, and legal boundaries.

Designs channel-specific go-to-market plans, asset checklists, sequencing, and kill criteria for a chosen offer or product. Use when the opportunity and offer are already chosen and the next question is how to reach buyers through the right channels instead of generic marketing advice.

Extracts and structures a user's skills, assets, access, constraints, credibility, and usable versus restricted edge into a reusable founder profile. Use when planning businesses, ranking opportunities, validating founder-market fit, or whenever later skills need a clean profile instead of raw conversation.

Standardize application error handling—taxonomy, propagation, response shapes, logging, and correlation IDs—when the task is to improve consistency of existing error behavior across a codebase.

Synthesizes validation experiments, weighs signal quality, identifies what actually changed, and recommends the next best experiment or decision. Use when a user has run multiple tests, has messy mixed evidence, or needs continuity across validation cycles instead of rethinking everything from scratch.

Guide selection and customization of Google Cloud Foundation Fabric or FAST Terraform modules when the task explicitly targets cloud-foundation-fabric patterns, factories, or landing-zone stages.

Decides whether a validated wedge should become software, defines the narrow product wedge, and separates what stays manual versus automated. Use when a service, workflow, or audience-driven opportunity may be ready for micro-SaaS scoping or when a user is tempted to build software too early.

Add production observability—health checks, metrics, tracing, alerts, and runbooks—when the task is to instrument or operationalize an existing service.

Implement or debug NIP-46 remote-signer (bunker/Nostr Connect) integrations when the task involves client-signer communication, bunker URIs, auth challenges, or signer daemons.

Implement or debug core Nostr client architecture—relay pools, subscriptions, deduplication, optimistic publishing, caching, and reconnect logic—when building or fixing a Nostr client.

Implement Nostr cryptography workflows—NIP-06, NIP-44, NIP-46, NIP-49, or NIP-59—when the task centers on keys, encryption, gift wrap, or remote signing in a Nostr app.

Implement NIP-90 Data Vending Machine clients or services when the task involves DVM job requests, feedback, results, payments, chaining, encrypted params, or NIP-89 discovery.

Construct valid Nostr event JSON and tag structures from requirements when the task is to create or validate concrete events rather than choose which NIPs to use.

Design correct Nostr REQ filters when the task is to query relay data with precise AND/OR, tag, thread, zap, reaction, or feed semantics.

Implement Nostr marketplace flows using NIP-15 or NIP-69 when the task involves stalls, products, auctions, bids, checkout messaging, or P2P orders.

Identify which Nostr NIPs apply to a feature and warn about deprecated choices when the task is protocol planning or NIP selection rather than concrete implementation.

Set up or troubleshoot NIP-05 identity verification when the task involves nostr.json, DNS/web hosting, CORS, redirects, or profile metadata for Nostr identity mapping.

Implement or extend a Nostr relay when the task involves WebSocket protocol handling, event validation, filtering, storage, subscriptions, or relay-side NIP support.

Implement Nostr social-graph and list features when the task involves follows, relay lists, mute/bookmark lists, NIP-51 lists, or outbox-model relay discovery.

Implement or debug NIP-57 zaps or NIP-61 nutzaps when the task involves zap requests, receipts, LNURL-pay, zap splits, Cashu tokens, or recipient payment configuration.

Turns a chosen opportunity into a concrete sellable offer with ICP, scope, deliverables, pricing hypothesis, and rejected framings. Use when a user has a best bet or validated direction and needs a clear commercial offer before selling or building.

Generate pull-request descriptions from branch diffs when the task is to explain change scope, motivation, testing, risk, and reviewer guidance for a reviewable branch.

Use when preparing, submitting, finishing, or cleaning up a pull request, especially when draft PRs, Greptile feedback, review comments, or merge-readiness work need end-to-end handling.

Create or iteratively improve agent skills with eval-driven refinement when the task is to build a new SKILL.md package or tune an existing skill’s trigger accuracy and performance.

Guards a project's telos/purpose hierarchy before scope, architecture, dependency, integration, roadmap, or implementation decisions. Use when the user asks should we build this?, does this fit?, where should this live?, what design aligns?, or when work may change product purpose, user experience, contracts, modules, or tests. Reads TELOS.md, telos/TELOS.md, telos/specs, and @telos annotations; initializes missing Telos context; evaluates L4 Purpose, L3 Experience, L2 Contract, and L1 Function before implementation.

Create secure GitHub Actions workflows for Terraform or Terragrunt plan/apply pipelines when the task is CI/CD automation for infrastructure deployments.

Converts a chosen opportunity into a fast, falsifiable validation plan with numeric targets, success criteria, and kill thresholds. Use when a user has a selected opportunity, best bet, offer, or product wedge and needs to test demand before building or scaling.

Use the Google Cloud gcloud CLI for GCP inspection, debugging, or administration when the task explicitly involves gcloud commands or Google Cloud resources; require confirmation before mutating resources.

Generate conventional commit messages from staged git changes when the task is to name or format a commit, not to review code or write release notes.

Operate the bundled PDF scripts to extract, OCR, create, merge, split, or convert PDFs when the task explicitly involves PDF document processing.