with one click
agentic-os
agentic-os contains 61 collected skills from coilyco-flight-deck, with repository-level occupation coverage and site-owned skill detail pages.
Skills in this repository
Kai's "start a sidequest" command - file a Forgejo issue from a described piece of work, then dispatch it through ward into its own session.
Tier and prune an issue backlog - P0-P4 tier definitions (P4 is the icebox/lowest tier), a target distribution as ranges, percentile-cut assignment that lands within the bands, pruning, and a second orthogonal automation-mode axis (headless/interactive/consult). Triggers - prioritize, prioritization, triage the backlog, P0/P1/P2/P3/P4, backlog ratio, icebox, burn down the backlog, tier the issues, automation mode, eligibility to dispatch.
gpg-ssm is a GPG wrapper that pulls signing material from AWS SSM at sign time instead of caching on disk. Use when wiring up a fresh machine for signed commits, bootstrapping the shared signing key, rotating a key, or debugging signing failures. Triggers - gpg-ssm, gpg signing, signed commits.
Git + GitHub umbrella. Hard rule - never use gh api graphql without double-confirming, REST is default. Owns the full PR lifecycle (branch, commit, open, monitor CI, auto-fix, merge) and routes to ward-passthrough and git-workflow siblings.
git workflow, forgejo, default tracker, commit, push, PR, gh, fj, issue, todo, gish.
Kai's technical preferences - Go CLI defaults (urfave/cli), Charm TUI stack, dead-repo check, doc conventions, default license. Triggers - cli, tui, library, dependency, recommend tool.
The five-tier information model for agent-facing CLI commands - description, body, intro, help, outro - plus push-short/pull-long. Use when documenting a CLI agents invoke, not humans.
Cookie-driven Playwright fallback for reading Honeycomb data when the official Honeycomb MCP server can't answer. Triggers - honeycomb ui scrape, honeycomb playwright, MCP fallback.
Inbound half of the scout pair. Capability gap analysis for skills and MCP servers - what to ADD. Sweeps repos, hydrates against skillsmp and glama, security-audits silver/gold tiers, installs one-issue-one-commit. Paired with tooling-scout-displacement (what to SHED).
Discipline for CLI security boundaries (primarily ward). Names anti-signals and load-bearing properties for designing, evaluating, and maintaining the boundary.
Author, modify, and validate agent skills. Covers naming, description caps, why-encoding, flat layout, Python helpers.
Warp is Kai's terminal on every host. Block-mode shell with left-side vertical tabs, one shell per tab. Preview is the Mac daily driver, Stable is the fallback.
Audit third-party packages, libraries, plugins, MCP servers, or upstream repos before pulling them in. Confirms maintainers real, code not malicious, project maintained.
Code review for ward's in-container review agent. Use when asked to review a filesystem diff, do adversarial review, same-family review, non-iterative review, or a ward review gate. Triggers - code review, review this diff, diff review, adversarial review, same-family review, non-iterative review, filesystem diff review.
Kai's licensing policy - MIT for shareable, AGPL-3.0 for deployment-of-one, proprietary for private. Triggers - license, LICENSE, MIT, AGPL, GPL, copyright, SPDX, what license.
Build leg of the scout family - what novel software to BUILD on your existing toolset, ranked for a long autonomous run, biased to refactors and wide backfills. Triggers - autonomy scout, scout autonomy, what should I build.
Run Tailscale alongside a second host VPN by containerizing it - userspace/SOCKS5 mode in Docker so SSH and tailnet reach coexist with a WireGuard-based commercial VPN. Use when two VPNs fight over the macOS route table, or wiring Tailscale SSH to a peer behind a host VPN. Triggers - tailscale, tailnet, vpn coexistence, tailscale ssh, two vpns, wireguard utun conflict, ward-tailnet, mac-proxy socks5.
Lint prose against Kai's voice rules - no em-dashes, no italics, no semicolons, no prose tables, she/her. Triggers - voice lint, voice check, lint prose, em-dash check, pronoun check.
Color and visual output for the Claude Code Desktop client by exploiting the fenced-block syntax highlighter (no ANSI). Triggers - render, colorful, color, terminal colors, chart, sparkline, ASCII art, braille art, visualization, syntax highlight trick, diff coloring, chat art, glyph palette.
AWS umbrella skill. Kai's primary cloud. Python via boto3, secrets and config via SSM Parameter Store, region pinned to us-east-1.
Umbrella for Go work across repos. Carries Kai's defaults (Go 1.22+, urfave/cli over cobra/kong, Charm stack for TUI, slog for logging) so agents inherit them before reaching for generic Go.
Kubernetes umbrella skill. K3s on kai-server is the homelab cluster. kubectl via ward wrapper. Helm for packaged apps. Plain manifests over Kustomize.
Category umbrella for building CLI tools. Go via urfave/cli, Python via click, Node via commander, shell prompts via gum. Wrapper APIs mirror the real CLI.
Category umbrella for building terminal UIs. Charm stack is the default - bubbletea, lipgloss, gum, glow, huh. Fall back to tview/promptui only when a project already commits to them.
Category umbrella for building HTTP / REST / GraphQL servers. Python via FastAPI (Flask on legacy), Go via stdlib net/http or chi, Node via Fastify. Async I/O, OpenAPI, Prometheus from day one.
Voice persona for ElevenLabs audio output - a Cortana-coded observer-narrator, voice id in SSM. Triggers - elevenlabs, tts, text-to-speech, voice generation, audio output, cortana.
Persona archetype for voice/text surfaces where the speaker is a passive observer-narrator, not an active agent. Constraints - observer not actor, single audience never named, passive voice past or future tense, neutral and unopinionated, methodologically specific. Triggers - persona, voice persona, narrator, observer.
Category umbrella for observability work across vendors - metrics, traces, logs, dashboards, alerting, SLOs. Current edge is OpenTelemetry plus LLM consumers of observability data.
Lazy MCP discovery via mcporter. Hard-trigger before any curl, gh, or HTTP fallback.
Kai's land-flow shorthands. `gish` lands one repo's current working tree (issue + commit minus lockdown files + push to Forgejo main). `gulp` drains everything else that could be lost (stashes, all branches, all worktrees) across every coily*/* repo - see references/gulp.md. Triggers - gish, land this, gulp, gulp every coily repo.
Harness profile for Aider, a git-native pair-programming CLI floating over swappable models (cloud or local Ollama). Triggers - aider, aider-chat.
Harness profile for Claude (Claude Code, Claude-family runtimes). Pronouns she/her. Capable cloud agent, large context, semantic skill-selection. The escalation target small local models hand up to. Triggers - claude, claude code, claude-code, anthropic agent.
Agent/harness profile for Codex (OpenAI's coding agent on GPT-family cloud models). Pronouns he/him. Capable cloud agent, peer to Claude, an escalation target. Triggers - codex, openai codex.
Harness profile for Goose, Block's extensible agent floating over swappable models, cloud or local Ollama. Triggers - goose, block goose.
Harness profile for OpenCode - a TUI/CLI harness floating over swappable models (local Qwen today). Pronouns they/them. Capability ceiling is the active models-* profile, not the harness. Triggers - opencode, qwen-opencode.
Discover and install skills from skillsmp.com (the Claude skill marketplace). Route through this skill, never raw curl - the harness denies direct skillsmp.com fetches.
Azure umbrella skill. Tertiary cloud - reach for it when a job already runs there or when Azure OpenAI is specifically required. AWS is Kai's default for new infra.
C# / .NET umbrella skill. Kai's first-class C# is Eco game modding against Eco.ReferenceAssemblies. Triggers - c#, csharp, .cs, .csproj, dotnet, nuget, msbuild, eco mod, EcoModKit.
GCP umbrella skill. Secondary cloud - reach for it when a job already runs there. AWS is Kai's default for new infra.
Kai cross-platform dotfiles - nushell config, in-process SSM secrets loader, shared aliases. Triggers - agentic-os, agentic-os-public, aos, in-process-ssm-loader, public-dotfiles, shared-aliases, zsh-config