Coding & Development

Review manufacturing-ready mechanical part models, engineering drawings, and assemblies. Covers ISO fits and tolerances, general tolerances (ISO 2768), ISO 20457 plastic molding tolerances, GD&T (ISO 1101 / ASME Y14.5), tolerance stack-ups, and process-specific DFM for injection molding, sheet metal fabrication & cutting, machining, additive manufacturing, and progressive stamping. Use when a user asks to inspect, review, assess, or find issues in a drawing, part, model, or assembly; check tolerances, fits, bearing fits, datums, tolerance stack-ups, or manufacturability.

Reviews one design document for completeness, internal consistency, implementability,

Designs and reviews MCP servers that agents use safely on multi-tenant products: tools as an adapter on the service layer, one tenant per connection bound in the token, rights checked per call, effect classes with preview and confirmation, replay protection, kill switch, error shapes, model evals. Use when building or reviewing MCP tools or servers. Not for using MCP servers or REST design.

Operate an OANDA fxTrade account through the `oandactl` CLI (OANDA REST-V20 API) - check balances, NAV and margin, list instruments, fetch prices, candles and order/position books, stream prices or transactions, and (only when the user explicitly asks) place, modify, cancel or close orders, trades and positions. Use when the user asks about their OANDA account, forex/CFD market data from OANDA, or trading on OANDA.

Use this skill whenever the user asks to set up, add, check or fix monitoring, alerts, uptime checks or a Watcher for this project — 'monitor this', 'watch my app', 'page me when it breaks', 'why didn't the alert fire', or extending an existing .watcher/ setup. It drives the `watcher` CLI end to end: init the project, scan it for telemetry points (and read the code for the silence a scan cannot see), add sources and one channel, prove every check can go red with `watcher doctor --live --prove`, run the server, and host it locally or on Cloudflare.

Primary pipeline execution step after /prd-to-issues or for clearly scoped implementation work. Use to build, verify, and commit a concrete slice, delegating to /tdd for backend work and behavior-heavy frontend logic when red-green-refactor will reduce risk. Not for shaping or pre-merge review.

Use when an agent is preparing or performing an OpenRig CLI/daemon upgrade and must preserve live seats, verify each mutation, or reconcile managed plugin and skill files without overwriting local work.

Primary pipeline review step after verified implementation. Use to create a PR with lineage and run architectural review before merge. Not for QA intake, planning, or implementation work.

Resumes interrupted work from transcripts, artifacts, memory, and git state without repeating completed actions.

Design and write the model-backed steps of a 247-agent workflow: the `decide` action (typed questions to TypeSafe's Jev classifier via OpenRouter's Decisions API, probabilities back, no text), the `llm` action (one Claude call with a JSON output schema) and the `agent` action (one session on an ACP agent such as claude-agent-acp, in a fresh git worktree, with tool-kind and command allowlists, a tool-call cap, a budget, a RESULT.json contract with `status: done | blocked` and deterministic `post` gates). Use it whenever a 247-agent task needs to classify, route, triage, extract, summarise, decide, or edit a repository; whenever the user mentions models, prompts, `system_file`, `output_schema`, `questions`, `criteria`, confidence thresholds, Jev, `max_tool_calls`, `budget`, `effort`, cost, tiers (Haiku/Sonnet/Opus), ACP, worktrees, \"blocked\" results or \"let an agent do X\"; and when implementing or reviewing the `llm`/`decide`/`agent` runners in `packages/core/src/actions/`. Do not skip it for \"just a quick

Opens, inspects, creates, updates, previews or reviews the active implementation plan (<prefix>-implementation_plan-<suffix>.md). Bare /plan returns links; /plan <instructions> drafts or updates, then self-reviews before presenting; /plan preview reopens in the host preview panel; /plan --review [suffix] audits a plan or design proposal before any code is written (architecture and convention fit, edge cases, security and authorization, data and migration safety, test coverage) with a READY / NEEDS REVISION / BLOCKED verdict, read-only until /proceed. Use --review for "review this plan", "does this plan make sense?", "poke holes in this design".

Find evidence-backed channel, audience, and product opportunities using Reddit, TikTok, and public aggregate ChatGPT usage data. Use to discover niches for new digital products or refine an existing product's positioning, workflow, features, or acquisition strategy.

Verify code changes in any repository with Spyglass. Use after changing behavior, when a claim needs proof, or when reviewing a verification run.

Runs a Lead-Worker workflow for code audits and delegated refactoring. The Lead (reviewer/orchestrator) scans code read-only, writes a gap analysis, drafts a whitelisted spec with a Definition of Done, dispatches a Worker subagent to implement it inside an isolated git worktree, gates the result (scope + DoD), and reviews the PR diff. With several tasks it runs Workers in parallel — all at once when the tasks are independent, or in dependency-ordered stages (parallel within a stage, integrated between stages) when they are related. Use when the user asks to audit a codebase and fix what is found, to delegate a refactor or cleanup to one or more subagents in separate worktrees, to run a spec-driven or Lead-Worker implementation, or to review a worker's branch before merging.

Prove a Finite feature works in the real app and keep the evidence. Launches an isolated headless run (Rust server, static web app, headless Chrome, MCP agents), drives the browser and the MCP tools the way a person and an agent would, and saves transcripts and screenshots to .verify/<run>/. Use before saying a change to the canvas, design list, comments, tokens, pages, import/export or an MCP tool works, before saying a bug is fixed, or when asked to verify, prove or demo one.

Open this branch's pull request and see it through the review loop from the author's side: the four gates one command at a time, the review record as the body, master merged in, the auto-merge label, then the wait each stage needs and the one manual step each stall needs, and just tidy after the merge

Model-check how Finite servers that share a Postgres store hand a design between them (leases, fencing tokens, the save task, MCP edits) with the TLA+ model in crates/server/tla. Use when a change touches Hub::tick, let_go, drop_copy, discover, persist, DesignHandle::load/retire/end/take_save/edit/unload_if_idle, Members::lapsed, or the Store claim/save/append/release/held calls, or when asked whether an edit can be lost when a design moves between servers.

Control Govee lighting through a capability-safe MCP server and Agent Skill. Use when an agent needs to inspect devices, discover scenes, compile a read-only plan, apply or restore a LightScore with explicit write gates, diagnose unavailable effects, author optional perceptual palettes, or install one managed lighting tool for Codex, Claude Code, and Hermes.

How to create an agent in Nemia — who it is (a name, a role, a description), its file and keys, its instructions, its closeout schema, its skills (and a third party's, vendored and credited), its fence, how it talks with a person in rounds, and how to check it loads and runs. For a person or a session writing or changing an agent, and for a review of one.

The first CLI for Afi SaaS backup - full public-API coverage plus the fleet-wide coverage, staleness, and offboarding answers the rate-limited API can't serve live. Trigger phrases: `check afi backup coverage`, `which mailboxes aren't backed up in afi`, `afi fleet health`, `offboard a user from afi backup`, `afi backup stale check`, `use afi`, `run afi-cli`.

smith check, smith fmt and smith lint: checking a project before committing, formatting with .clang-format, clang-tidy with .clang-tidy, and on .smith files. Load before committing, when a check fails, or on 'code should be clang-formatted' and clang-tidy findings.

Use this skill when the user wants to: (Mode 1) Design and package a complete SAP Cloud Integration (CPI / Integration Suite) iFlow + Script Collection OSGi .zip bundle from a natural-language functional specification using the 3-stage pipeline (Plan -> BPMN2 .iflw XML -> Groovy Scripts), (Mode 2) Reverse-engineer and generate comprehensive 8-section technical documentation from an existing iFlow .zip bundle, OR (Mode 3) Clone and adapt an existing reference iFlow and Script Collection .zip bundle to a new API specification.

Manage engineering work with native harness subagents as Workers, cost-conscious model selection, and HAPI synchronization between Managers across sessions and machines.

Write the technical spec at docs/specs/<slug>/_techspec.md — explore the architecture, settle technical decisions one question at a time, record them as ADRs, and produce a build order that write-tasks can decompose. Entry point for feature work with an approved PRD, and also for refactors and bug fixes, where it mints the spec folder with a minimal _prd.md instead of requiring a product interview.