kaos

Fix every open Dependabot PR end-to-end on autopilot. Use this skill when asked to run /dependabot-fix-all (no arguments). This skill acts as an orchestrator that discovers all open Dependabot PRs once, risk-orders them, then processes them one at a time by spawning an isolated non-interactive `copilot -p` child per PR that runs the dependabot-fix skill. It verifies each PR independently via gh, applies the merge policy, records state in a durable ledger, and is bounded so it always terminates. Never pauses for user input.

Comprehensively diagnose and fix a failing Dependabot PR. Use this skill when asked to run /dependabot-fix <pr-number>. The user provides the PR number in their prompt. The skill loads PR context, surveys errors at a high level, ingests relevant repo instructions / docs / source via subagents, performs a deep root-cause diagnosis, designs a risk-tiered fix with a manual testing strategy, commits the fix directly to the Dependabot PR branch, posts a REPORT.md as a comment on it (never commits it), and evaluates whether the skill itself needs updating afterwards.

Plan and execute complex KAOS implementation work with staged context gathering, backend/UI synthesis, validation, PR/CI checks, and REPORT.md PR-comment output.

Execute a full KAOS release. Use this skill when asked to release a new version of KAOS. The user provides the target version (e.g., v0.5.0) in their prompt. This skill handles pre-flight checks, release creation, CI monitoring, artifact validation, smoke testing, and post-release cleanup.