jelbirt
- 23 skills
- 0 followers
- 6 hours ago last updated
- ▌
- ▌ Commit · jelbirtCreate a git commit for the pending changes, auto-generating a message in the repo's commit style and evaluating the actual diff (not just file names) to decide whether it should land as one commit or be split across several. Use whenever the user says "commit this", "commit these changes", "commit it", "make a commit", or otherwise asks to commit — even without typing /commit explicitly. Applies the "would I revert these together?" test to catch unrelated concerns (mixed features, refactor + feature, unrelated code areas) and always proposes the commit plan for approval before staging or committing anything.
- ▌ Digest · jelbirtProduce a cross-project status digest - a short, decisions-first brief synthesized from each repo's worktrees, branches, open PRs, CI state, and todo/state docs, ordered by what needs the reader's attention. Use whenever the user wants a status digest, weekly brief, Monday brief, standup summary, or cross-project status, says "digest", "weekly digest", "status brief", "what moved last week", or "where does everything stand", or when a scheduled digest run fires. Read-only - it never commits, writes files, or mutates any repo.
- ▌ Handoff · jelbirtEmit a successor-session briefing from the CURRENT session - exact state left behind, in-flight and promised-but-undone work, open owner gates, honest uncertainties, standing invariants, and orient-first instructions - in one fenced code block. Use whenever the user wants to wrap up, hand off, or continue in a fresh session, says "handoff", "write a handoff", "brief the next session", "I'm about to /clear", or asks what a successor needs to know, even if they don't say "handoff" explicitly. The audience is a successor that starts empty - use session-report instead when reporting back into a session that is still running and already has the project in context.
- ▌ Kickoff · jelbirtGenerate a complete, copy-ready kickoff prompt for a new Claude Code session from a short workstream description - orient-first checks, inherited state, ordered work items, hard stops, and definition of done, returned in one fenced code block. Use whenever the user wants to spin up, kick off, or brief a new session or workstream, says "kickoff", "write a kickoff prompt", "prep a session for X", or "hand this plan to a fresh session", even if they don't say "kickoff" explicitly.
- ▌ Sidebar · jelbirtAnswer a tangential side question without bloating the main session - delegate it to a subagent that researches it in its own context (the repo first, reaching out to the web only when the answer isn't in the project) and returns only a tight, cited answer. Use whenever the user asks a side/aside question mid-session, says "sidebar", "side question", "quick aside", "while we're here", "remind me how X works", or wants to look something up about the project's architecture, design, or existing/planned code or infra without spending main-chat context. Prefer this over answering inline so the heavy file-reading stays in the subagent and only the distilled answer returns.
- ▌ Git Undo · jelbirtDiagnose a git mistake and return the safe steps to undo it - the exact commands, what each does, what is reversible, and how to verify. Use whenever the user wants to undo, revert, roll back, or recover something in git, such as an accidental or wrong commit, committing to the wrong branch, an unwanted reset --hard, a bad merge or rebase, a force-push that lost work, a deleted branch, unstaging files, or "I think I lost a commit." Prefer this over ad-hoc git advice so recovery is state-aware and destructive steps are flagged and snapshotted first. If the repo is not directly accessible, ask the user to paste their git status and git reflog output.
- ▌ Rename To Match Folder · jelbirtOne sentence on what this does, then when to use it with natural trigger phrases a user would actually say. Lead with the key use case. Lean slightly pushy - skills under-trigger. Max 1024 characters.
- ▌ Audit Fix · jelbirtImplement fixes for findings from the preceding /audit report, parallelizing via subagents grouped by non-overlapping file scope. Accepts optional severity/category filter.
- ▌ Flowchart · jelbirtGenerates a clean, readable flowchart of a codebase, feature, component, process, or entire project and renders it inline in the chat window. Accepts an optional argument describing the scope (e.g. "the auth flow", "the entire repo architecture", "how requests move through the API layer") plus any other specs like orientation or detail level. Use whenever the user asks for a flowchart, flow diagram, architecture diagram, process diagram, or wants to visualize how a feature/component/system/process works - e.g. "diagram this", "flowchart the X flow", "show me how X works visually", "draw the architecture", or "/flowchart". Delegates codebase exploration to a subagent and renders via the visualize tool to keep token usage low.
- ▌ New Skill · jelbirtScaffold a new Agent Skill - interview for intent, then create the folder and a conventions-following SKILL.md with valid, pre-filled frontmatter, and lint it. Use whenever the user wants to create, start, scaffold, or bootstrap a skill, says "new skill", "make a skill for it", "turn this into a skill", or is about to hand-write a SKILL.md. Prefer this over writing a SKILL.md from scratch so the name, description, and layout follow conventions. For heavy eval or benchmark-driven iteration, hand off to skill-creator instead.
- ▌ Deps Audit · jelbirtProduce a unified, prioritized report on project dependencies across three dimensions - security vulnerabilities, version freshness, and unused or missing packages - by orchestrating npm/yarn/pnpm audit, outdated, and depcheck into one triaged action list. Use when the user wants a dependency audit, asks if packages are outdated, wants to check for vulnerabilities, wants to find unused dependencies, or wants a dependency health report - this goes beyond automated security-only alerts like GitHub Dependabot by also covering freshness and utilization.
- ▌
- ▌ Next Steps · jelbirtFigure out the logical next steps for the current project by delegating a token-efficient review to a subagent that reads the repo state, git history, and the most recent plan/roadmap docs, then returns a prioritized short list. Use whenever the user asks "what's next", "what should I work on", "next steps", "where did we leave off", "what's left", "help me plan the next session", or seems unsure how to move the project forward — even if they don't say "next-steps" explicitly.
- ▌
- ▌ Math Research · jelbirtReview a mathematical or algorithmic choice against the literature - delegate it to a research-only subagent that reads the actual implementation, identifies which branch of applied math the question sits in, and returns a cited findings brief with complexity bounds, numerical-stability analysis, and tradeoffs. Use whenever the user asks whether an algorithm is the right one, wants the state of the art for a numerical method, questions convergence, conditioning, or stability, asks for a paper or textbook citation to justify an approach, wants alternatives to a solver, estimator, decomposition, or embedding compared, or says "math research", "is this the right method", or "what does the literature say". The dividing line against a quick sidebar lookup is citation - use this whenever the answer must trace to a primary source rather than to recall, however plausible the recall sounds.
- ▌ Release Notes · jelbirtGenerate release notes from git history since the last tag, grouped into features, fixes, breaking changes, and improvements using conventional-commit prefixes. Use when the user wants release notes, a changelog, or a summary of what shipped since the last release - works directly from raw commit history, so it fits direct-to-main workflows or repos that don't rely on GitHub's PR-label-based release notes generator.
- ▌ Scaffold Audit · jelbirtRun the periodic audit of Claude Code scaffolding - inventory every skill, command, agent, hook, permission list, and model pin across ~/.claude, triage each against the decay patterns that make old-model-era components harmful on newer models, gather A/B bake-off evidence for the contested ones, then prune with verified reference cleanup and git-versioned rollback. Use when the user says "scaffold audit", "prune my config", "audit my skills/setup", "Cherny prune", or "delete your CLAUDE.md and see what happens", when a new Claude model generation ships, or roughly every 6 months since the last audit.
- ▌ Session Report · jelbirtWrite the status report this session sends back to the main session that delegated the work - the scope it was given, what it accomplished, exactly what changed on disk, how that was verified, what remains outstanding, and how the remainder connects to the tasks the receiving session is managing - in one fenced code block. Use whenever a delegated, side, or parallel session has to report up, or the user says "session report", "report back", "write up what you did", "status synopsis", "summarize this session for the main session", or "what do I paste back". Written for a reader that already has the project in context and merges into a session that keeps running, so use handoff instead when the next session starts empty.
- ▌ Prompt Engineer · jelbirtDesign, diagnose, or refine a prompt for Claude or another LLM using a standalone prompt-engineering framework - clarify the goal and audience, structure role, context, task, and format, apply techniques like few-shot examples and structured output, and diagnose why an existing prompt underperforms. Use whenever the user wants to write a system prompt, user prompt, tool description, or agent instructions from scratch, wants to fix a prompt that isn't working well, or asks for help crafting, refining, or optimizing a prompt for any LLM.
- ▌ Prune Test Suite · jelbirtReview the test suite for relevance against current source code, pruning stale tests and adding missing coverage
- ▌ Worktree Cleanup · jelbirtTear down finished git worktrees safely - verify each branch is merged and its tree is clean, then remove the worktree, delete the branch, and prune. Sweeps every merged-and-clean worktree at once, or targets one branch. Use when the user wants to clean up, tear down, close out, or remove a worktree or workstream, says "worktree cleanup", "sweep merged worktrees", "remove finished worktrees", or asks to delete a finished branch's directory. Delegates to the repo's own scripts/rm-worktree.sh when one exists, and otherwise runs the equivalent gated steps directly.
- ▌ Sync Profile Readme · jelbirt bundleRegenerate the project table in a GitHub profile README from your pinned repositories, so repo descriptions stay the single source of truth and the profile stops needing hand-maintenance. Use whenever the user wants to sync, refresh, or update their profile README, says "update my profile README", "my pinned repos changed", "regenerate the projects table", or "my profile table is stale", or has just pinned, unpinned, published, or re-described a repo. Reads pins through the GitHub GraphQL API (the REST API does not expose pinned items), rewrites only the region between the PROJECTS:START and PROJECTS:END markers, never rewrites a hand-written row, never deletes a row, and always shows a diff for confirmation before writing.