AI & ML

delorenj Bundle

Develop pjangler itself: author Commands (atomic file/dir operations), Recipes (composed subsystem bootstrappers), and register them in the CLI. Covers the Command/Recipe architecture, project registry implementation, CommonProject copier implementation, and pjangler dist/build/regression workflows. Use when creating a pjangler Command or Recipe, registering a recipe, adding subsystem bootstrapping, changing templates/commonproject or templates/hermes-agent, debugging pjangler tests, or changing the CLI/MCP server. Triggers: pjangler command, pjangler recipe, add subsystem, bootstrap, project scaffolding, CommonProject template, hermes-agent template, pjangler CLI, pjangler MCP. Do NOT use for: USING pjangler to create a new project (→ projects); generic agent-config fan-out engine mechanics (→ agent-config-fanout); versioning many files in parity (→ mise-versioning); event schema naming (→ bloodbank-integration).

delorenj Bundle

Autonomous TDD development loop with parallel agent swarm, category evolution, and convergence detection. Use when running autonomous game development, quality improvement loops, or comprehensive codebase reviews.

delorenj Bundle

Guidance for creating, running, fixing, and promoting behavioral evaluations. Use when verifying agent decision logic, debugging failures, debugging prompt steering, or adding workspace regression tests.

delorenj Skill

Find focused, runnable Deepgram recipes for a specific feature × language. Use whenever someone wants a minimal working code snippet for ONE feature (transcribe URL, diarize, smart-format, voice agent connect, etc.) rather than a full starter app. Recipes are under 50 lines, read DEEPGRAM_API_KEY from env, and ship with a runnable example_test. Covers Python, JavaScript, Go, .NET, Java, Rust, and the Deepgram CLI.

delorenj Bundle

Operate PJangler Project Notebook as bounded, Git-grounded project memory. Use when an agent needs to inspect notebook status, read or update the stable Overview, manage scoped notes, search project knowledge, audit or migrate a binding, recover capture receipts, or install/check the canonical global Project Notebook hooks without changing foreign hooks.

kevinnft Skill

HuggingFace hf CLI: search/download/upload models, datasets.

kevinnft Bundle

Axolotl: YAML LLM fine-tuning (LoRA, DPO, GRPO).

x402agent Skill

Convert desired token amounts to base lots — the most common agent mistake.

delorenj Bundle

Agent-local context guard and deterministic compaction workflow (GOD-13). Use when context usage is high, when wiring auto-compaction/restart timers, or when deploying standardized context-monitor hooks across agents.

delorenj Bundle

Use when building or maintaining a personal LLM-powered knowledge base. Triggers: ingesting sources into a wiki, querying wiki knowledge, linting wiki quality, 'add to wiki', 'what do I know about', or any mention of 'LLM wiki' or 'Karpathy wiki'.

jacob-balslev Skill

Use when deciding whether to route a task to the fast/cheap tier (Claude Haiku) — transcription, polling, format conversion, structured-output slot-filling, small-diff review, high-volume low-latency work — and where the boundary is that should escalate to Sonnet/Opus. Covers the cost/latency floor, the 200K context ceiling (vs the upper tiers' 1M), the absence of the effort knob, the separate rate-limit pool, and when to drop below Haiku to a script. Do NOT use for ordinary multi-step feature work (use claude-sonnet), the hardest reasoning (use claude-opus), loop design (use autonomous-loop-patterns), or Claude API request syntax (read the claude-api reference).

jacob-balslev Skill

Use when writing, tightening, evaluating, or repairing an LLM prompt or reusable prompt template for completion, agent dispatch, grading, structured extraction, tool use, or prompt-engineered workflows. Covers instruction hierarchy, message roles, context placement, few-shot examples, structured output, positive constraints, reasoning guidance, prompt-injection resistance, provider differences, and eval-driven iteration. Do NOT use for whole context-system design (use context-engineering), eval dataset or grader design (use eval-driven-development), reviewing generated code (use code-review), authoring SKILL.md files (use skill-scaffold), choosing which skill or agent should activate (use skill-router), or root-causing a deployed failure after outputs already exist (use debugging). Do NOT use for review this AI-generated PR for correctness. Do NOT use for scaffold a new skill that teaches prompt engineering.

jacob-balslev Skill

Use when deciding whether to route a task to the balanced implementation tier (Claude Sonnet) — feature work, bug fixes, test writing, multi-step code — as the default lane that is cheaper/faster than the frontier tier and more capable than the fast tier. Covers the cost/quality tradeoff vs Opus and Haiku, the shared 1M context window, effort behavior, and the 1M-context subscription billing caveat. Do NOT use for the hardest reasoning/architecture/security work (use claude-opus), high-volume mechanical or low-latency work (use claude-haiku), loop design (use autonomous-loop-patterns), or Claude API request syntax (read the claude-api reference).

jacob-balslev Skill

Use when designing or auditing the multi-graph context architecture of an AI-coding workspace: skill graph, document routing graph, memory index, script registry, and the cross-graph edges between them. Covers edge typing, orphan detection, connectivity health, deterministic graph synthesis signals, change-propagation checks, and drift or hub-and-spoke anti-patterns. Do NOT use for authoring one SKILL.md (use `skill-scaffold`), validating one skill (use `skill-infrastructure`), live routing decisions (use `skill-router`), context-window budgeting (use `context-window`), or session load/drop choices (use `context-management`). Do NOT use for scaffold a new SKILL.md from a template. Do NOT use for validate that this single skill's frontmatter matches the schema. Do NOT use for decide which skill to inject for this query right now. Do NOT use for this skill says 'use orgQuery'; that one says 'never use orgQuery' — fix the conflict.

jacob-balslev Skill

Use when reasoning about how a system, user, or designer's internal model of behavior may diverge from reality — applies across UX, distributed systems, type systems, API design, and team collaboration. Covers the three-model frame (designer / system image / user), the two gulfs (execution and evaluation), analogy and metaphor as model-seeding, the five failure modes (transfer, overgeneralization, underspecification, drift, invariant blindness), the surface/operational/architectural/domain layering, and the discipline of validating a model against the system it claims to represent. Do NOT use for the visual representation of a model (use knowledge-modeling), for the formal-domain entities-attributes-relationships of conceptual modeling (use conceptual-modeling), for cognitive biases in decision-making (out of scope), or for empirically eliciting user models via research methods (use user-research). Do NOT use for name the React hook for managing form state (tactical implementation choice).

kevinnft Skill

Interactive QA session where user reports bugs or issues conversationally, and the agent files GitHub issues. Explores the codebase in the background for context and domain language. Use when user wants to report bugs, do QA, file issues conversationally, or mentions "QA session".

kevinnft Bundle

Outlines: structured JSON/regex/Pydantic LLM generation.

delorenj Skill

Set up the Deepgram MCP server for your AI coding tool. Checks whether the Deepgram CLI (dg/deepctl) is installed: if so, uses the local CLI MCP server (dg mcp) for full tool access; otherwise offers the hosted documentation MCP or suggests installing the CLI. Use whenever someone wants to install Deepgram's agentic tools, set up the MCP server, or connect their editor to Deepgram.

kevinnft Bundle

Control a running TouchDesigner instance via twozero MCP — create operators, set parameters, wire connections, execute Python, build real-time visuals. 36 native tools.

kevinnft Skill

Decomposition playbook + specialist-roster conventions + anti-temptation rules for an orchestrator profile routing work through Kanban. The "don't do the work yourself" rule and the basic lifecycle are auto-injected into every kanban worker's system prompt; this skill is the deeper playbook when you're specifically playing the orchestrator role.

kevinnft Bundle

Integrate TinyFish web toolkit (search, fetch, browser automation) into Hermes Agent

jacob-balslev Skill

Use when condensing prose while preserving meaning: session findings, wrap reports, research briefs, executive summaries, TLDRs, agent handoffs, progressive summaries, audit summaries, and long-document distillation. Covers extractive vs abstractive summarization, what to keep vs drop, evidence preservation, summary levels, handoff summaries, and audit-report condensation without hiding findings. Do NOT use for byte/data compression algorithms (use `compression`), context-window budget math or compaction triggers (use `context-window`), working-set selection (use `context-management`), prose tone repair (use `writing-humanizer`), or quality scoring (use `evaluation`).

jacob-balslev Skill

Use when allocating context-window budget across system, skill-injection, working, and output zones; monitoring context health; deciding when to compact; preserving state before compaction; recovering after compaction; or choosing strategies for 1M, 200K, or 128K context windows. Covers zone budgets, practical model-budget tables, the 80% compaction rule, pre/post-compact protocols, persistence hierarchy, operation token costs, and token-reduction techniques. Do NOT use for deciding what information belongs in the working set (use `context-management`), prompt design (use `prompt-craft`), graph architecture (use `context-graph`), or memory curation. Do NOT use for decide what context to load or drop in the working set. Do NOT use for design the multi-graph architecture for skills + docs + memory. Do NOT use for improve the prompt template the agent uses. Do NOT use for curate the durable memory index across sessions. Do NOT use for which skill should activate for this query.

jacob-balslev Skill

Use when discovering a domain through events, commands, actors, policies, aggregates, read models, external systems, and temporal workflows before implementation. Do NOT use for event schema/topic contracts (use `event-contract-design`), webhook handler implementation (use `webhook-integration`), generic state transition modeling (use `state-machine-modeling`), or persistence schema design (use `entity-relationship-modeling`). Do NOT use for implement Shopify webhook signature verification and idempotent retries. Do NOT use for draw the state machine for this one status field. Do NOT use for create a normalized data model and indexes. Do NOT use for write event-bus infrastructure code. Do NOT use for define the schema, topic, compatibility, and fixtures for a selected event.

jacob-balslev Skill

Use when deciding whether to spend GitHub Copilot's metered budget on a task (its premium-request / post-June-2026 AI-credit model, where 1 credit = $0.01 and cost = tokens × per-model rate), what Copilot is good at vs expensive at, which plan allowance (Pro 300 / Pro+ 1500, no rollover) applies, and when a cheaper or free lane should take the work instead. Covers the June 1 2026 shift from premium-request multipliers to usage-based token billing, the always-free completions/next-edit surface, and the IDE-native frontier-model lane. Do NOT use for choosing or operating the OpenCode runtime (use `opencode`), for picking a specific free model (use `opencode-free-models`), or for authoring an agent loop (use `autonomous-loop-patterns`). Do NOT use for how do I invoke the opencode CLI non-interactively? Do NOT use for which free model fits this bulk job? Do NOT use for how do I write the agent's retry loop?

jacob-balslev Skill

Use when creating a new SKILL.md from scratch, restructuring a draft before it becomes a stable skill, or teaching another author the canonical Skill Metadata Protocol frontmatter, body, and audit-state.json sidecar structure. Covers flat schema-conformant frontmatter, the sidecar split, v8 classification, body layout by skill intent, semantic-layer discipline (description vs activation vs Coverage), teaching-layer mechanics (TEMPLATE NOTE blockquotes), native skill-creator handoff, public/private safety, and routing-eval honesty. Do NOT use when modifying an already-written skill (edit it directly), writing general technical documentation, routing an existing request across skills, or fixing malformed library health at scale (use `skill-infrastructure`). Do NOT use for refactor my existing skill to be more concise. Do NOT use for my skill's routing isn't activating — why? Do NOT use for audit my skill library for stale frontmatter. Do NOT use for write a developer guide for the contributor docs.

delorenj Bundle

Keep one hand-edited master config and generate per-agent CLI configs from it. Covers SSOT fan-out for agent hooks and skills: master-to-dialect propagation, ambiguity lock files, drift checks, skill packs, and the Bloodbank services/agent-hooks reference implementation. Use for hooks.master.json, hooks.mappings.lock.json, generated-config drift, fan-out, SSOT, agent hooks, inherit_global, .agents/skills.json, sync.py, sync-skills.py, provision-packs.py, skills-provision-packs, project-scoped hooks, skill fan-out, and new agent CLI dialects. Also use for skill packs: packs[], pack.toml, SHA256SUMS, sealed/immutable packs, payload verification, pack version resolution, include/exclude, and the six supported CLI skill dirs. Do NOT use for using pjangler to create projects or adoption checklists (33god-projects), event schemas or Bloodbank topology (bloodbank-integration), versioning (mise-versioning), or single-target config.

delorenj Bundle

Control and operate OpenCode for coding work using the local OpenRouter-only preset setup. Use this skill to manage OpenCode sessions, select one of the approved model aliases (openrouter/fusion, openrouter/3-buck-chuck, openrouter/free-lunch), switch Plan/Build agents, and coordinate implementation through OpenCode.

kevinnft Bundle

OBLITERATUS: abliterate LLM refusals (diff-in-means).

kevinnft Skill

Webhook subscriptions: event-driven agent runs.

kevinnft Skill

Discovers and invokes agent skills. Use when starting a session or when you need to discover which skill applies to the current task. This is the meta-skill that governs how all other skills are discovered and invoked.

kevinnft Bundle

Triage issues through a state machine driven by triage roles. Use when user wants to create an issue, triage issues, review incoming bugs or feature requests, prepare issues for an AFK agent, or manage issue workflow.

jacob-balslev Skill

Use when reasoning about the protocol-level cycle by which a language model uses external tools: declaration, request, execution, continuation, model-visible transcript or provider-resumed state, ID pairing, tool-result formatting, vendor encodings (Anthropic Messages, OpenAI Responses and Chat Completions, MCP, Gemini generateContent and Interactions), client tools vs hosted/server tools, strict schemas and grammar-constrained custom tools, streaming arguments, tool search/deferred loading, programmatic tool calling, computer-use/browser-control screenshot-action loops, parallel vs sequential calls, error handling, and the separation between model intent and runtime execution. Do NOT use for deciding when or how many tools to call (use tool-call-strategy), multi-agent architecture (use agent-engineering), prompt wording (use prompt-craft), or eval design for tool-use behavior (use eval-driven-development).

jacob-balslev Skill

Use when moving slow or failure-prone work out of a request path, designing job queues, retries, checkpoints, progress reporting, cancellation, or worker concurrency. Covers inline-vs-background decisions, queue contracts, state machines, idempotency, retry/backoff, progress signals, worker leases, and user-visible completion reporting. Do NOT use for time-based schedule design (use `cron-scheduling`), live browser transport choice (use `real-time-updates`), or async message schema ownership (use `event-contract-design`). Do NOT use for choose the cron expression for a daily run. Do NOT use for design an SSE or WebSocket browser update channel. Do NOT use for define an event envelope and topic naming standard. Do NOT use for debug why this already-running worker crashed. Do NOT use for model the database schema for the business entity being processed.

jacob-balslev Skill

Use when a team is converging on solutions before agreeing on the problem, when a brief reads as a feature request, when symptoms and root needs are tangled, or when assumptions need surfacing before design work proceeds. Do NOT use for code-level bug triage, runtime failure diagnosis, or root-cause analysis of system errors — those are engineering investigation tasks, not design problem framing. Do NOT use for Find the bug causing the 500 error in the checkout endpoint. Do NOT use for Why is the test suite flaky on CI? Do NOT use for Classify whether this agent request is high-risk before executing.

delorenj Bundle

Set up aggressive recurring project execution loops (“Crank Engines”) using cron + agent turns, with a strict finish contract. Use when the user asks to keep cranking on a project/sprint (e.g. “set up an aggressive crank engine”, “keep pushing until sprint is done”, “run status every X minutes”), and needs configurable project, recipients, focus scope, quantifiable completion criteria, screenshot evidence, and manager-go/no-go enforcement.