Detects a dispatched agent gone silent mid-run from events.jsonl. Use during multi-agent runs or when a run feels quiet — silent agent, stalled run, heartbeat check, timeout.
Art direction for mobile app reference screens. One screen per image, portrait device aspect, subtle straight-on phone frame, status bar with realistic content, real-looking…
Add GPU-driven indirect drawing with compute-based frustum culling to an SDL GPU project. A compute shader tests bounding spheres against frustum planes and writes indirect draw…
Initialize or repair a repository for the Forge workflow. Use when a project has no docs/status.md, when the user asks to set up Forge, or when Forge project artifacts need to be…
Forge playbook for business automation and API integrations — Gmail, Calendar, CRM, webhooks, Slack, payments. Use for integration, OAuth, sync, notification, automation.
Streaming LLM output correctly. SSE setup, partial-token UTF-8 decoding, AbortController cancellation end-to-end (browser + server + upstream), mid-stream errors as events not…
Structured logging discipline. JSON-first, correlation IDs, level meaning, redaction at source, sampling, never-log-secrets, audit logs separated from application logs.
Forge doctrine for using MCP servers as a client — opt-in, tiered, least-privilege. Use when a Boss needs an external tool grant or when installing/discussing an MCP server.
Exposing resources via MCP. URI scheme design, listing strategies, mimeType discipline, ranged reads, mutable resources via subscribe, when resources beat tools, multi-tenant…
Designing an MCP server LLM agents actually use well. Transport choice (stdio vs HTTP streaming), capability scoping, structured error responses the model can recover from, tool…
Designing individual MCP tools that LLMs use correctly on the first try. Naming, descriptions written for model consumption, JSON Schema discipline, error returns that the model…
Forge playbook for legacy modernization without a big-bang rewrite. Use for migration, legacy, port, upgrade, strangler fig, characterization test, parity, feature flag, cutover.
Forge playbook for mobile apps — React Native/Expo/Flutter. Use for mobile app, iOS, Android, Expo, EAS, Flutter, apk, ipa, secure storage, deep link, App Store, Play Store.
Naming variables, functions, types, files, tables. Ban list for AI defaults (`data`, `Manager`, `Helper`, `utils.ts`), domain-specific over abstract, length proportional to scope,…
Resume a project and choose the next normal action. Use at the start of a new session, after opening an unfamiliar repository, or whenever the user asks what to do next in a…
Overnight session-limit-resilient builder — work-package queue, auto-resume, morning briefing. Use for run this overnight, keep going while I'm away, resume tomorrow, opt-in only.
Production observability via OpenTelemetry. Three pillars (traces, metrics, logs) with explicit correlation, OTel SDK setup, span discipline (semantic conventions, error capture),…
Run one evidence-driven capability optimization cycle and accept/revert the candidate.
Forge playbook for payment/checkout/billing — Stripe, Mollie, Adyen, PayPal. Use for payment, checkout, subscription, refund, webhook signature, idempotency key, PCI, 3D Secure.
Web performance discipline. Core Web Vitals (LCP, INP, CLS), image optimization (modern formats + responsive srcset + lazy loading), JS bundle discipline (code-split +…
PR descriptions that get reviewed and merged fast. Title under 70 chars, Summary/Why/Test plan structure, screenshots for UI changes, risk callouts for sensitive areas, length…
Forge playbook for prediction, sports-data, and betting analysis (incl. Telegram tips). Use for forecast, odds, value bet, backtest, confidence label — NO real-money betting.
Generates a top-tier project-specific CLAUDE.md. Use when a project has no CLAUDE.md or a thin/generic one — project brain, onboarding, project identity, hard rules, governance.
Retrieval-augmented generation discipline. Semantic chunking with hierarchy, hybrid retrieval (BM25 + dense + rerank), citation discipline, refusal paths, eval-driven iteration,…
Infrastructure reconnaissance — inventory all cloud resources, map connections, flag risks. Use when asked to "inventory our infra", "what infrastructure do we have", "map our…
Master agent for Update Pipeline stage ② (Reconcile). Reads the context bundle from a reconcile_run row, reasons over the change, produces one of four verdicts (no-op | apply |…
Upgrading an existing app or website to premium quality without breaking it. Audit first, identify the AI smell (indigo gradient, card-in-card, cramped spacing), surgical edits…
Builds the global, opt-in, read-only Forge Project Registry across all projects. Use when the owner wants a bird's-eye view of status, last run, tickets, and ports.
PROJECT-LOCAL OVERRIDE for jarvis-agents (TBD). Fires at status=released after forge-test has verified the branch locally.
Use when the user wants a repository-specific UI or design skill synthesized from current agent skills; not for ordinary UI implementation or backend-only work.
Forge V2 router — classifies a task by domain/complexity and selects the smallest fitting agent team. Use at the START of any /forge run or build/create/automate request.
Create a persistent expert agent file. Use when "create an agent", "make an agent", "design an agent", "forge an agent", "agent for [task]", "I need a specialist", "build a [role]…
Relational schema design. UUIDv7 primary keys, foreign-key enforcement at the DB, audit columns, soft delete with partial unique indexes, enums vs lookup tables, multi-tenancy…
Forge playbook for SAFE, legal web scraping and data collection. Use for scrape, crawler, spider, harvest, extract data, API pull, lead list — public/permitted sources only.
Add a procedural anti-aliased grid floor to an SDL3 GPU scene using fwidth()/smoothstep() in the fragment shader.
Create a new SKILL.md from scratch and iterate with forge eval until it passes the quality gate. Don't use for editing an existing skill — use forge-skill-improver instead.
Turn one recurring workflow from supplied or available conversations into a focused, evaluated skill candidate.
Improve an existing SKILL.md with forge eval until it passes the quality gate, fixing the weakest category each round.
Meta-guide for Forge's per-project pipeline skills: what Forge does with a skill at runtime, and the standard for authoring / customizing / shadowing one correctly and cheaply.
Regenerates FORGE_SNAPSHOT.md from real project sources so no session loses the mission across compaction. Use on /forge snapshot or let PreCompact/SessionStart hooks trigger it.
Premium expensive UI register. Warm neutrals, refined typography (Söhne / Inter Display / GT America), generous spacing, considered shadows, spring motion.
Exports sealed Obsidian vault notes into soul_traces.jsonl format for HuggingFace persistence. Implements snapshot isolation, git pre-flight checks, and consumes the…
Evaluating subagent outputs before trusting them. Shape and budget validation, refusal handling, partial-result recovery, hallucination defenses (URL/quote/file verification),…
LLM function/tool calling discipline. Tool selection, parameter design, structured returns, error-as-data, parallel tool calls, iteration caps, context budget across turns,…
Schema-first input validation at every trust boundary. zod / valibot / pydantic, derive-types-from-schema, structured field-level error responses, env validation at startup, file…
Runs Forge's verify-loop — checks a completion claim against events.jsonl and the ticket store, sends it back if false. Use after every agent/work-package completion in a run.
Adversarial verifier for Update Pipeline stage ② (Reconcile). Independently judges whether a reconcile agent's candidate skill body is safe to publish, using only the bundle from…
Forge playbook for AI voice/phone agents — Dutch-first calls, turn-taking, safe handling. Use for voice agent, voicebot, phone, IVR, Twilio, Vapi, ElevenLabs, TTS, STT, barge-in.
Forge safe-parallel-work playbook using git worktrees. Use when more than one Boss writes code in the same repo at once — parallel agents, worktree, isolate, avoid conflicts.
บุคคลปลอมแปลงเอกสารราชการ เอกสารสิทธิ์ หรือเอกสารสำคัญ เช่น โฉนดที่ดิน ใบขับขี่ บัตรประชาชน หนังสือเดินทาง หรือลงลายมือชื่อปลอม แก้ไขข้อความ ตัดทอน ประทับตราปลอม…
Use when the user wants Codex to work with the Elephant Hand Forgejo instance, Forgejo or Gitea repositories, issues, pull requests, organizations, users, releases, Actions…
ForgeRock integration. Manage data, records, and automate workflows. Use when the user wants to interact with ForgeRock data.
Metapromptore Generativo di Entità Semantiche PERMANENTI (Forgia v1.0). Attivare quando l'utente richiede 'crea agente', 'genera skill', 'nuovo agente', 'crea entità', 'genera…
Use when a task needs the judgment of a Forging Machine Setter, Operator, or Tender — verifying actual billet temperature at the point of forging rather than trusting furnace…
Trigger Pattern Always (run during recon TASK 0, not breadth) - Inject Into Recon agent only (meta_buffer.md enrichment)
Walk a contributor through forking a GitHub repo they don't own, branching, committing, and opening a PR back to upstream – using `gh` and `git`.
Keep fork-decoupled workers strictly bounded by directive text and cache sharing requirements.
L1 trigger - audits fork-choice rule implementation (LMD-GHOST, Tendermint locking, Nakamoto longest-chain) for equivocation handling, slot-vs-block reasoning, duplicate block…
Create a development branch tied to a GitHub issue using a consistent naming convention.
Fork divergence monitor - tracks where the fleet's active forks diverge in CODE (unique commits, new/modified skills) and CONFIG (enable/var/model/schedule vs upstream), gated on…