Atypical-Consulting

Fix failing GitHub Actions pipelines directly — detects broken workflows, reads run logs, diagnoses root causes, applies fixes in worktrees, and creates PRs. No prior repo scan or backlog required. Use this skill whenever the user wants to fix CI, repair a pipeline, debug Actions failures, or says things like "fix my pipeline", "fix CI", "my actions are failing", "fix my workflows", "repair my build", "CI is broken", "fix GitHub Actions", "debug my workflow", "why is my build failing", or "action fix". Do NOT use for scanning repos (use ghs-repo-scan), applying backlog items (use ghs-backlog-fix), or merging PRs (use ghs-merge-prs).

Applies backlog item fixes to a GitHub repository using parallel worktree-based agents — clones the repo once, creates git worktrees for each fix, launches agents simultaneously, verifies acceptance criteria, and creates PRs. Uses dependency-aware wave execution to prevent conflicts between related items and recalculates health scores after each wave for incremental progress. Persists session state for multi-session continuity. Use this skill whenever the user wants to apply, fix, or resolve a backlog item, references a project item, or says things like "apply this backlog item", "fix this issue", "resolve this finding", "work on this backlog item", "apply tier-1--license", or references an item in a GitHub Project. Also trigger when the user says "apply all backlog items", "fix all findings", "resolve all tier 1 items", or "apply all for {repo}". Do NOT use for scanning repos (use ghs-repo-scan), viewing backlog status (use ghs-backlog-board), or general code review.

Implement a GitHub issue using worktree-based agents, then create a PR. Clones the repo, creates a worktree, spawns an agent to implement the fix/feature, verifies the result, and opens a PR with auto-close references. For complex issues (High/Very High complexity from ghs-issue-analyze), automatically uses GSD framework for multi-phase planning and execution with wave-based parallelization, fresh context per task, and automated verification. Use this skill whenever the user wants to implement an issue, fix a bug from an issue, build a feature from an issue, or says things like "implement issue #42", "fix issue #42", "implement #42", "build feature from issue #15", "implement all triaged issues", "implement all bugs", "work on issue #42", "code issue #42", or describes work with no issue number like "owner/repo refresh readme" or "owner/repo fix the broken badge". Do NOT use for triaging/labeling issues (use ghs-issue-triage), analyzing issues (use ghs-issue-analyze), applying backlog health items (use ghs-ba

Shows a dashboard of all backlog items across all audited repositories — health scores, open issues, progress, and what to work on next. Use this skill whenever the user wants to see backlog status, check progress across repos, find the next item to fix, or asks things like "show backlog", "what's left to fix", "backlog status", "dashboard", "show all repos", "remaining items", "how are my repos doing", or "progress report". Also trigger when the user says "list backlog", "show findings", "show issues", or just "backlog". Do NOT use for scanning new repos (use ghs-repo-scan), applying fixes (use ghs-backlog-fix), or reviewing code. For quick "what should I work on next?" queries, prefer ghs-backlog-next instead.

Recommends the highest-impact next backlog item to fix across all audited repositories. Use this skill whenever the user wants a quick recommendation, asks "what should I work on next", "next item", "what's the most important fix", "highest impact item", "what to fix next", "priority item", "next action", or "recommend something to fix". Also trigger for "what's next" or just "next" in the context of backlog work. Do NOT use for full dashboards (use ghs-backlog-board), scanning repos (use ghs-repo-scan), or applying fixes directly (use ghs-backlog-fix).

Calculates and displays the health score for a repository from its backlog items. Use this skill whenever the user wants to see a repo's health score, check their score, recalculate after fixes, or says things like "what's my score", "show score for {repo}", "health score", "recalculate score", "how healthy is my repo", "score check", or "points breakdown". Do NOT use for full dashboards with issue lists (use ghs-backlog-board), scanning repos (use ghs-repo-scan), or applying fixes (use ghs-backlog-fix).

Promotes draft health items in a GitHub Project to real GitHub Issues for team visibility. Creates issues from draft items in the Todo column with Source = Health Check, replaces drafts with linked issues in the project, and copies custom field values. Use this skill whenever the user wants to sync health findings to GitHub Issues, promote draft project items to real issues, create issues from health checks, push findings to GitHub, publish health items for the team, or says things like "sync backlog", "create issues from health checks", "push findings to GitHub", "publish backlog", "sync health items", "sync findings to issues", "promote drafts", or "create GitHub issues from scan". Do NOT use for scanning repos (use ghs-repo-scan), applying fixes (use ghs-backlog-fix), or viewing backlog (use ghs-backlog-board).

Act as an autonomous developer for a repository — monitor for new/triaged issues, process them through the full development lifecycle (triage, analyze, implement, review, merge), handle health maintenance, dependency updates, and optional releases. Operates in cycles with human checkpoints for safety. Use this skill whenever the user wants a full developer loop, says things like "dev loop", "be my developer", "autonomous developer", "work on my repo", "process all issues", "developer mode", "full dev loop", "maintain this repo", "handle everything", "automate development", "dev cycle", "work through issues", "run a dev cycle", "be my dev", or "take over this repo". Do NOT use for single-skill operations — use the individual skill directly. Do NOT use for multi-repo pipelines — use ghs-orchestrate instead.

Use when the user wants a comprehensive code review from multiple perspectives, or invokes a team-based review combining business, design, and engineering viewpoints with debate and consensus. Trigger this skill whenever the user asks for a "legends review", "team review", "full review", "review from all angles", or wants business + design + engineering perspectives debated together. Also trigger when the user wants Elon, Jobs, AND Linus to review the same code, or asks for a multi-perspective review with debate and consensus. This is the flagship skill — use it when no single persona is enough.

Use when the user requests a code review with technical and business recommendations. On-demand review combining Elon Musk's first-principles thinking with structured diff analysis, observability, and broader architecture assessment. Trigger this skill whenever the user asks for an "elon review", "elon style", "musk review", "first principles review", "business value review", or wants to know if a feature should exist, if the code could be 10x simpler, what to delete, or what the opportunity cost is. Also trigger when reviewing architecture decisions, velocity concerns, or anything where both technical rigor AND business context matter.

Use when the user requests a code review focused on design quality, developer experience, and product vision. On-demand Steve Jobs persona review evaluating API elegance, error messages, documentation, simplicity, and whether the product deserves to exist. Trigger this skill whenever the user asks for a "jobs review", "jobs style", "steve jobs review", "design review", "DX review", "developer experience review", or wants to know how their API feels, whether the naming is elegant, if the onboarding is smooth, or if the product has taste. Also trigger when reviewing error messages, documentation quality, or anything where the human experience of using the code matters.

Use when the user requests a code review focused on engineering correctness, code quality, and technical rigor. On-demand Linus Torvalds persona review evaluating abstractions, error handling, security, logging/observability, performance, and whether the code is actually good. Trigger this skill whenever the user asks for a "linus review", "linus style", "engineering review", "correctness review", "technical review", deep code quality analysis, or wants to know if their code is actually correct, thread-safe, or maintainable. Also trigger when reviewing error handling, DI patterns, race conditions, or anything where rigorous engineering scrutiny is needed.

Autonomously burn down a GitHub issue backlog with a FLEET of N parallel sub-agents — each takes one issue through its whole lifecycle (`ai-migration-kit:implement-issue` → `ai-migration-kit:merge-pr` → report) then picks up the next, so N issues are always in flight. The orchestrator ABOVE `ai-migration-kit:implement-issue`/`ai-migration-kit:merge-pr`/`ai-migration-kit:create-issue`: surveys open issues, orders them (small effort first, then medium), assigns each worker a NON-overlapping code area to avoid conflicts, keeps exactly N running (retires a worker the moment its PR merges and dispatches a replacement), has workers file off-scope finds as fresh issues, and verifies real merge state itself (workers tend to idle at "PR ready" without landing). ALWAYS use it when the user wants to work through MANY issues hands-off with parallelism — "implement issues small first then medium with 3 agents", "keep N agents continuously implementing and merging issues", "burn down the backlog", "run the auto-dev loop",

Turn a task or instruction into Socratic prompts — questions structured as theory → framework → application that make the model surface its own quality criteria and reason before producing output, instead of jumping straight to a generic answer. Use whenever the user wants to write, design, or upgrade a prompt this way, and especially when they mention "Socratic prompt(s)", "Socratic prompting/method", turning an instruction into questions, making a prompt "think/reason first", or getting sharper, less generic LLM output — including loose phrasings like "rewrite this prompt so the AI reasons first", "give me N Socratic prompts about X", "make this prompt sharper", or pasting an instruction-style prompt and asking to upgrade it. Does NOT cover actually executing the underlying task — only crafting the prompt(s).

Launch and drive Koine Studio in WEB mode — the browser host of the Studio IDE (`tooling/koine-studio` run as a plain web page with the browser backend + Blazor WASM compiler, no Tauri/Rust sidecar). Use when asked to run, start, serve, screenshot, or verify a change in Koine Studio Web / the WASM studio in a browser. For the Tauri DESKTOP build (`npm run tauri dev`) or the `Koine.Cli` compiler, this is the wrong skill.

Use when the user wants to write, draft, or create an Architectural Decision Record (ADR) — capturing an architecture or technical decision, recording why a choice was made, or documenting a decision for the team. Guides framing a sharp Context/Decision/Consequences and drives the AdrMcp MCP tools (create_adr, validate_adr) to produce a well-formed MADR record. Triggers on "write an ADR", "record this decision", "create a decision record", "document this architecture choice", "we decided to…".

Use when the user wants to review, critique, or sanity-check an Architectural Decision Record (ADR) — an existing record or a draft. Evaluates whether the decision is sound, falsifiable, honestly scoped, and well-linked, using the AdrMcp MCP tools (get_adr, validate_adr, detect_conflicts, find_related_adrs). Triggers on "review this ADR", "critique my decision record", "is this ADR any good", "check ADR-000X", "does this decision hold up".

Use when an existing architectural decision is being changed, replaced, retired, reversed, or deprecated — the team changed its mind and a prior ADR no longer reflects reality. Drives the correct lifecycle transition and bidirectional linking via the AdrMcp MCP tools (supersede_adr, set_status, link_adrs, find_related_adrs). Triggers on "supersede ADR-000X", "we changed our mind about", "replace/retire this decision", "this ADR is out of date", "mark as deprecated/superseded", "reverse decision".