21StarkCom
Official@21StarkCom
21StarkCom publishes Claude Code and Codex skills for PR review, Terraform/Terragrunt audits, GitHub Actions cost reduction, release management, and long-form editing.
Agent Skills by 21StarkCom
Showing 26 vetted skills indexed across 1 GitHub repositories.
21StarkComstark-gha-cost
Diagnose and reduce GitHub Actions and GHAS billing across organizations and enterprises.
21StarkComstark-blog-sharpen
Diagnoses and edits long-form blog posts to cut padding and remove AI writing tells.
21StarkComstark-release
Automates changelog generation, semver version bumps, git tagging, and GitHub Release creation.
21StarkComstark-voice
Drafts Slack messages and notes in Aryeh's Hebrew-English code-switching voice.
21StarkComstark-handover
Persists session state as numbered handover chains and progress trackers for cross-session resumption.
21StarkComstark-review-improvement
Applies targeted edits to code review prompts and orchestrator based on assessment findings.
21StarkComstark-build
Implements an accepted spec autonomously with fresh writers, deterministic gates, and one advisory review.
21StarkComstark-author
Authors human-gated specs and task DAGs through recon, interview, and sign-off.
21StarkComstark-jury
Runs one post-editing skill across a three-model panel and reconciles the results.
21StarkComstark-logging
Reviews and writes structured application logs with correct levels, fields, and correlation.
21StarkComminion
Implements assigned coding tasks in isolated worktrees and reports progress to a coordinating leader.
21StarkComstark-init-docs
Scaffold, backfill, migrate, and clean standardized developer documentation structures in git repositories.
21StarkComstark-session
Manages development session start briefings and end-of-session test, merge, commit, and push workflows.
21StarkComstark-story-judge
Grades long-form blog posts with cold cross-vendor LLM judges on an anchored rubric.
21StarkComstark-persona
Assigns a weighted-random character persona voice to the agent for the session.
21StarkComstark-review
Reviews GitHub pull requests across triage-selected domains using a single configurable LLM agent.
21StarkComstark-fresh-eyes
Dispatches one zero-context reviewer to verify document claims by a different method.
21StarkComstark-refactor-plan
Inspect a repository and generate an evidence-based refactor plan and machine-readable backlog without modifying source code.
21StarkComstark-adr
Create, list, and supersede MADR-lite Architecture Decision Records in a repository's docs/adr directory.
21StarkComgru
Coordinates multi-worker ticket execution with dispatch, verification, and recovery in Codex sessions.
21StarkComstark-terragrunt-review
Reviews Terragrunt orchestration HCL across multiple LLM agents for dependency, state, and include issues.
21StarkComstark-story-edit
Rewrites long-form drafts into narrative posts while preserving every fact byte-identical.
21StarkComstark-bury
Retires code subsystems or entire repositories into a versioned graveyard with gated deletion PRs.
21StarkComstark-terraform-review
Reviews Terraform and OpenTofu HCL code across multiple LLM agents for security and correctness findings.
Frequently Asked Questions About 21StarkCom
FAQPage SchemaWhat tasks can I accomplish with 21StarkCom's skills?▼
You can review PRs and Terraform/Terragrunt code with multi-model panels, reduce GitHub Actions and GHAS billing, cut releases with changelogs and tags, plan refactors, manage ADRs, scaffold docs, and edit long-form posts in a specific personal voice.
Who are these skills built for?▼
They target software engineers and engineering managers working in Claude Code or Codex runtimes — particularly Aryeh Kiovetsky's team — who need PR review, infrastructure auditing, CI cost control, release management, and voice-matched writing assistance.
How do the review skills run in practice?▼
Skills like stark-review, stark-terraform-review, and stark-terragrunt-review dispatch claude, codex, and gemini as separate subagents, then merge and cross-validate findings. They are review-only by default; posting, fixing, and committing require explicit opt-in flags.
Are 21StarkCom's skills free to use?▼
The skills are published as open skill manifests in public repositories with no stated license fee. Runtime cost depends on the underlying model usage, since most skills specify the opus model and some dispatch multiple vendor models in parallel.
What prerequisites do these skills require?▼
Most skills require a Claude Code or Codex runtime environment and a git repository context. Specific skills need extra setup: stark-memory and stark-design-tokens run only on claude, gru and minion run on codex, and stark-adr wraps the `brain adr` command.