Vorbereitung mit Pre-mortem und Grenz-Trockenlauf, Gate vor dem Start, nachts Risiko zuerst mit Stop-Hook, Wartebank statt Abbruch, Meldungen per md-Datei oder ntfy, geschützte Richter-Dateien, Morgenbericht mit „Jetzt du“ oben.
Mercury banking for Claude: connects to the hosted Mercury MCP server (OAuth) and adds finance skills for spend analysis, anomaly detection, and cash reconciliation.
Workflow tooling: wrap-up report (artifact, Bear, Slack), mental-model pages, MCP server audit, Claude config audit, local-LLM setup advice, and a hook that wakes a session waiting on a background task about once an hour for a one-line reply keeping the prompt cache warm (on by default, CACHE_KEEPALIVE=off to disable)
Research → design → plan → implement → review pipeline. Orchestrator skills plus the reviewer, researcher, and implementation agents they delegate to, and core engineering principles.
Analysis & estimation suite (enabled by default): rfp-requirement-analysis → rfp-architecture-design → rfp-sa-bdd forms a traceable RFP/SOW pipeline; cold-estimation delivers method-based initial estimates with risk effort for new and existing systems; itemized-estimate turns a locked estimate into a one-page customer-facing itemized sheet in the customer's own Excel template; quality-orchestrator runs a writer→reviewer→fixer checklist loop.
Drive PicPrep from your assistant - choose which photos make the cut, arrange the keepers into the running order of a post, frame landscapes for a portrait-first feed, watermark, and export at a platform's size - handing the judgement calls back to a person instead of guessing at them.
Youssef Mostafa's Claude Code skills in one place: plan-build-board (approval-gated planning with a live status board, tasks grouped into reviewable jobs, a PRs tab with independent and stacked PRs and copyable configs, an Issues tab tracing issues to tasks, jobs and PRs), integration-assistant (git merges planned up front: parallel branch analysis, semantic-conflict detection, a tabbed decision review, and a review loop that rebases each stacked PR and resolves Claude's review on it), handoff (verified, cumulative session handoff docs), debug-log-sink (ephemeral localhost log collector with verified cleanup), resolve-pr-reviews (PR review comments from AI reviewers or people verified, fixed test-first, answered per thread and closed with one PR comment) and sha3bolly (junior-engineer explanation style).
Open-source AI job search that runs locally in your AI coding CLI: scan job portals, evaluate listings into a structured A-H report with a 1-5 score, tailor your CV, track applications
Open-source AI job search that runs locally in your AI coding CLI: scan job portals, evaluate listings into a structured A-H report with a 1-5 score, tailor your CV, track applications
Turn a big task into a tree of agents. A routing skill decides when to decompose work, picks the right fan-out pattern (deep-review, parallel-research, design-tournament, migrate), and drives reusable Workflow scripts with planner/worker/verifier/synthesizer agent archetypes.
Session discipline: verify claims, push back, stress-test plans with operator and adversary personas, and grade deliverables 0-100 with a reviewer agent.