AI Agent Skills

This skill supports SAP S/4HANA migration projects including approach selection (brownfield, greenfield, bluefield), SAP Readiness Check analysis, simplification item re…

Captures the user's macOS screen, a window, or a region with the built-in `screencapture` CLI, then reads the image so the agent can see what the user sees. Use it when…

Set up local development linting by picking the right linters for each file type a project uses, then recording them as required pre-commit checks in AGENTS.md / CLAUDE.…

Read the most recent AI Diff Reviewer review from the current branch's open PR, present the findings in the same format as the parent skill's local review flow (verdict…

Author a well-documented GitHub pull request — title and body — for the current branch. Reads the diff and commit trail, infers a Conventional Commits (or repo-native) t…

Challenge, push back, play devil's advocate on AI output or intentions. Use when: challenge this, are you sure, push back, prove it, what if you're wrong, devil's advoca…

Safe-to-fail experiment for Complex domain problems where cause-effect is only visible in retrospect. Two-phase: foreground qualify → background probe → sense result. Us…

The Midwit skill is the user’s ongoing anti-complexity gate for a conversation. It triggers on /midwit or requests to maintain an anti-overengineering mode, keep the use…

The voice-mode skill owns the conversation’s delivery and interaction behavior when the user needs to listen instead of look at a display. It triggers on explicit “/voic…

Install, authenticate, and verify the geekbot CLI end to end. User-triggered via /geekbot:geekbot-setup.

Report the install + auth state of the geekbot CLI. Safe for Claude to invoke proactively before running Geekbot actions.

Read a number without being fooled by it: an exit code, a passing result file, a test count, a list size, a control run, a red or a green. Use before reporting any figur…

Decide whether a failing test run is the change's fault or was already broken on main, before retrying, reverting or blaming anyone. A red is a measurement, not a verdic…

Diagnose a failing test or reported defect, propose a fix, and apply it only with approval at each stage — investigate, then fix, then verify. Use for the same job as th…

Investigate software defects and unexpected behaviour using read-only evidence before remediation. Use when the user asks to diagnose, find a root cause, explain a failu…

Forces the laziest solution that actually works, simplest, shortest, most minimal. Channels a senior dev who has seen everything: question whether the task needs to exis…