myrgic
- 36 skills
- 0 followers
- 6 hours ago last updated
- ▌ Btw · myrgicFork the current session into a parenthetical aside using cog_fork_session. The aside runs in a child session with role "btw-aside" so it has its own context and does not pollute the parent thread. On completion, the parent session resumes. Use when the user types /btw or says "by the way" and wants to pursue a tangent without losing the main thread.
- ▌ Voice · myrgicDual-modal communication protocol for voice-enabled AI development. Guides when to speak vs write, non-blocking speech patterns, and anti-patterns. Requires mod3 at localhost:7860. Tools differ by platform — see Platform Context section.
- ▌ Pr Triage · myrgicEnumerate, classify, and summarize open pull requests across the myrgic GitHub org. Use when asked to triage PRs, check what needs review, audit open work, or surface stale or blocked PRs across repos.
- ▌ Codex · myrgicUse when the user asks to run Codex CLI (codex exec, codex resume) or references OpenAI Codex for code analysis, refactoring, or automated editing
- ▌ Handoff · myrgicHand off work between Claude Code sessions through the CogOS kernel bus — emit a handoff.offer with state blob when your context is near-full or the task is paused, or claim a pending offer in a fresh session to resume another session's work. Use when you're running out of context, when a task should continue across sessions, or when bootstrapping a new session from another session's state.
- ▌ Threads · myrgicRegister and check open-ended waits — "I'll hold until the verdict lands" — as a resolution predicate in the threads registry (~/.cog/status/threads.json), so the intention survives compaction and session end instead of dying silently in conversation context. Use whenever you state that you're waiting on something external (a CI verdict, a review decision, an operator reply, an async job) and the wait could plausibly outlive this conversation turn or this session. Also covers how to write a good predicate vs a bad one, and how to close a thread once it's actually been looked at.
- ▌ Council · myrgicConvene parallel deliberation agents that reason independently then synthesize. Use when you need multi-perspective analysis of a design decision, architectural choice, or complex problem. Spawns background agents with different lenses, coordinates via shared workspace artifacts, and produces a unified synthesis.
- ▌ Suggest · myrgicPause and produce a numbered decision menu — invariants, context, and 3-5 options the user can pick by number or override with a custom message. Use when the user types /suggest mid-conversation.
- ▌ Issue Triage · myrgicEnumerate, classify, and surface actionable open issues across myrgic repos. Use when the issue backlog needs review, before sprint planning, or when onboarding contributors.
- ▌ Consolidate · myrgicReflect on the current session's arc when context pressure or operator signals indicate an inflection point. Map settled vs heating-up distinctions, surface at-risk-of-being-lost items, capture any new operational framing as feedback memory, write a consolidation cogdoc + continuation brief. Use when context approaches ~60% saturation, when the operator signals stepping away, when a long arc has settled and needs durable capture, or when an explicit `/consolidate` is invoked. Distinct from `handoff`, which is transactional (one session emits, another claims) — consolidate is reflective (distill the session's distinctions into durable form).
- ▌ Cold Start · myrgicWrite a cold-start handoff document for a project, feature, or initiative. Creates a standalone briefing that allows any agent or human with zero prior context to understand the full picture and continue the work.
- ▌ Orchestrate · myrgic bundleDispatch a Sonnet master orchestrator that fans out N parallel Sonnet/Haiku workers, synthesizes their work, and returns to the Tier 1 parent. Use when ≥3 parallel bounded items need parallel execution + synthesis. Triggers on /orchestrate, "orchestrate this", "dispatch in parallel", "have agents work on this in parallel", "knock these out in parallel", "fan out workers".
- ▌ Plan Phases · myrgicProduce a phased execution plan for multi-step implementations. Decomposes complex tasks into dependency-wired task graphs with wave assignments, role mappings, verification gates, and parallel execution schedules. Use when a task has 5+ discrete steps or requires coordinating work across multiple files or domains.
- ▌ Execute Plan · myrgicExecute a phased plan produced by plan-phases. Transforms dependency-wired task graphs into running work with parallel subagents per wave, verification gates between waves, and persistent progress tracking. Use after plan-phases produces a plan and the user approves it.
- ▌ Local Review · myrgicReview a pull request using a local model (Ollama/vLLM) via pull-context dispatch. The model navigates the actual diff, reads relevant files, and produces a verdict grounded in real code — not in a flat summary blob. Use when you want a peer review from a local model without paying cloud inference tokens, or when you need a review before committing to a PR submission.
- ▌ Code Quality · myrgicProtocols for writing maintainable code - clarity, testing, error handling, refactoring, and best practices. Use when implementing features, reviewing code, or improving existing code.
- ▌ Retrospective · myrgicRe-examine earlier work through the lens of a new result, milestone, or discovery. Launches parallel analyst agents per domain, producing standardized retrospective reports that track what changed, what was validated, and what new questions opened.
- ▌ Git Forensics · myrgicDeep knowledge of git internals - object model, content addressing, hash verification, and repository inspection. Use when debugging git issues or understanding repository state at the byte level.
- ▌ Macos Storage · myrgic bundleAnalyze and manage disk storage on macOS. This skill should be used when the user asks about disk space, wants to free up storage, needs to find large files, clean caches, or manage developer build artifacts. Handles macOS-specific concerns like sparse files (Docker.raw), cloud storage placeholders (Google Drive streaming), and developer tool caches. Triggers on requests like "what's using my disk space", "free up storage", "clean caches", "find large files", "how much space do node_modules use".
- ▌ Dispatch Agent · myrgic bundleSpawn, manage, and collect results from subagents. In Hermes, primary dispatch is delegate_task; in Claude Code, primary dispatch is the Agent/Task tool. CLI backends (Codex, Claude Code, OpenCode) remain available for cross-agent work and isolation scenarios. Use when delegating work to subagents or orchestrating multi-agent workflows.
- ▌ Lab Engineering · myrgicLab execution protocols for computational experiments - environment verification, defensive coding, checkpointing, and reproducibility. Use when running experiments, validating environments, or ensuring results are reproducible.
- ▌ Critical Review · myrgicRigorous evaluation of arguments, theories, claims, and architectural decisions. Includes steel-manning, assumption hunting, scope validation, falsification analysis, and alternative explanations. Use when reviewing designs, proposals, or technical claims before committing to them.
- ▌ Openclaw Expert · myrgicExpert guide for OpenClaw's gateway hub-and-spoke architecture, 20+ messaging platform integrations, session management, agent execution, and multi-agent routing.
- ▌ Zsh Environment · myrgic bundleManage and troubleshoot the zsh shell environment on macOS with Oh My Zsh. Use this skill when the user asks about PATH problems, missing commands, shell configuration, oh-my-zsh plugins, shell startup performance, environment variables, exposed secrets in dotfiles, terminal application settings (VS Code, iTerm2, Ghostty), or any zsh/shell troubleshooting. Triggers on requests like "fix my PATH", "why is my shell slow", "add a plugin", "audit my shell config", "check for secrets", "configure my terminal", "what's in my PATH", "command not found".
- ▌ Session Protocol · myrgicProtocols for starting and ending work sessions - initialization, checkpointing, handoffs, and continuity. Use when beginning or ending a session, or creating handoff documentation.
- ▌ Experiment Design · myrgicDesign rigorous experiments with clear hypotheses, controls, metrics, and analysis plans. Covers A/B testing, ablation studies, benchmarking, and proof-of-concept validation. Use when planning experiments to test technical hypotheses or compare approaches.
- ▌ Inference Cascade · myrgicDesign and run supervised multi-agent inference cascades using idle local GPU capacity. Spawns parallel agents with read-wide write-narrow sandboxing, collects observations as append-only hash-chained deltas, and triangulates findings from model agreement and disagreement. Use when running autonomous research loops, evaluating model performance against a live corpus, or exploring a problem space with multiple models in parallel.
- ▌ Technical Writing · myrgicProtocols for effective technical documentation - structure, clarity, specification, and executor-oriented writing. Use when writing specifications, reports, proposals, or documentation.
- ▌ Physics Validation · myrgicProtocols for validating physical plausibility of computational results - dimensional analysis, limiting cases, order-of-magnitude estimates, and analytical cross-checks. Use when checking if results make physical sense.
- ▌ Literature Research · myrgicStructured literature review and source synthesis. Guides systematic search, source evaluation, claim extraction, and cross-reference analysis. Use when researching a technical topic, evaluating prior art, or building a knowledge base from published sources.
- ▌ Systems Architecture · myrgicProtocols for designing reliable systems - failure mode analysis, operational concerns, scalability, and infrastructure patterns. Use when designing systems that need to survive production.
- ▌ Pull Context Dispatch · myrgicDispatch bounded work to sub/peripheral agents by pointer-reference — pass identity + directive + tool access + accessible-state pointers, NOT recapitulated context or summaries. Use when delegating to subagents, authoring an agent/Task prompt, or deciding what context to hand an agent. The agent observes what it needs rather than receiving pre-summarized context.
- ▌ Substrate Archaeology · myrgicSurface, integrate, and map prior work in the CogOS substrate — find existing research, prior cartography, theoretical lineages, and connect them to current work before building new things.
- ▌ Corpus Cross Reference · myrgic bundleCross-check a proposed RFC or ADR against the existing substrate corpus. Surfaces composition gaps, conflicts, prior art, frontmatter validation, and numbering issues. Runs at proposal time — invoke manually today; auto-fires on RFC/ADR PR-open via a future GitHub Actions hook. Triggers on /corpus-cross-reference, "cross-reference the corpus", "check this RFC against existing", "what does the corpus already say about", or when opening an RFC/ADR PR and wanting a composition review.
- ▌ Interdisciplinary Research · myrgicCross-domain synthesis and connection-finding using analogical reasoning, boundary-spanning, and novel hypothesis generation by linking disparate fields. Use when exploring new ideas, finding unexpected connections, or synthesizing across disciplines.
- ▌ Kanban Closed Loop Supervisor · myrgicClosed-loop orchestration over Hermes delegate_task + Kanban. Use when >=3 parallel bounded items need parallel execution + synthesis AND a quality criterion must be satisfied before proceeding (reviewer must GO). Binds the harness-agnostic orchestrate pattern to Hermes dispatch machinery. Extends kanban-orchestrator; does NOT replace it.