indrasvat

Author a new Agent Skill (a SKILL.md plus optional references, scripts, and hooks) for any task or domain. Use when the user says "write a skill", "create a skill", "author a skill", "scaffold a skill", "new SKILL.md", "make a Claude Code skill", "turn this…

Drive ONE persistent, already-authenticated Chrome — the user's real logged-in sessions — over the DevTools Protocol, instead of spawning a fresh unauthenticated browser per task. Use when a task needs the browser SIGNED IN AS THE USER: screenshotting a…

Drive iTerm2 programmatically using Python scripts to automate terminal tasks, run visual tests, or manage sessions. Use when the user needs to build a CLI, build a TUI, create a terminal app, add features to a CLI or TUI, fix a bug in a CLI or TUI, test…

Detect breaking changes in APIs — protobuf, OpenAPI/Swagger, GraphQL schemas, or Go exported interfaces. Use when the user says "API compat", "breaking changes", "check API compatibility", "proto breaking", "schema breaking", "is this breaking", "backward…

Run all CI checks locally before pushing — linters, type checks, tests, and vulnerability scans in parallel with a pass/fail summary table. Use when the user says "run CI", "check before push", "run all tests", "lint check", "pre-push checks", "validate…

Analyze the impact of CRD (Custom Resource Definition) changes — find all controllers, operators, webhooks, RBAC rules, and manifests that reference the CRD and need updates. Use when the user says "CRD impact", "what breaks if I change this CRD", "CRD change…

Clean AI-generated slop from code — remove unnecessary comments, dead code, redundant logic, and rename for self-documentation. Use when the user says "clean this up", "deslop", "remove slop", "clean code", "remove unnecessary comments", "simplify", "code…

Create a self-contained execution plan (ExecPlan) for a feature, refactor, migration, or complex task. Produces a living markdown document any agent or developer can follow end-to-end without prior context. Use when the user says "create a plan", "write an…

Drive terminal sessions, panes, and TUIs from an agent, and prove what they render. Five distinct jobs — multiplex terminal work (sessions/windows/panes) as a tmux/screen replacement; snapshot any pane as a pixel-perfect PNG, headless, with no terminal…

Record a before/after demo video that shows the REAL product failing and then working — a live terminal session, not a script narrating its own verdicts. Use when a bug or fix needs to be *seen* rather than described: "make a demo video", "record a…

Spawn parallel adversarial subagents that DRIVE THE REAL SYSTEM (not reason from source) to break new code before it ships — schemas, contracts, parsers, serializers, protocols, state machines, security/redaction, freeze guards. Use as a standard step on any…

Structured GitHub PR state for AI coding agents. Use for PR readiness, CI failure diagnosis, unresolved review threads, bounded review waiting, bot-review triage, and safe stale-review dismissal. Trigger immediately after PR creation and again after every…

Orchestrates multiple AI coding CLIs (Claude, Codex, Antigravity, Grok) as a council. Use for a second opinion from another model, multi-perspective code review, peer review and synthesis across agents, validating a PRD or design for gaps, iterative…

Use gh-hound's structured GitHub Actions CI surface for agent workflows: inspect runs, detect failures, watch fail-fast, parse JSON failure triage (job, step, exit_code, log_excerpt), and branch on gh-hound exit codes instead of screen-scraping the TUI.

Run a realistic yathaavat debugger dogfood gauntlet with subagents: generate an isolated buggy Python/uv target app, have a separate persona debug it through the yathaavat TUI driven by shux, capture UX/DX friction, verify findings, and implement product…