This agent helps Python teams using the Anthropic or OpenAI SDK safely cut LLM spend without risking regressions. With zero code changes, it maps call sites and runs your test scenarios to rank token waste from missed prompt caching or heavy model tiers. It tests fixes in isolated git worktrees, enforcing a strict "do-no-harm" gate that reverts any patch failing to preserve behavior. It delivers a clear savings audit that converts into verified, behavior-preserving code patches upon opt-in.
Built from hands-on experience running AI agents on Android/Termux aarch64. Every command and workaround in this skill was discovered through real builds, not docs. Tested across multiple devices and model sizes.
Reliability patterns for running three or more Claude Code agents unattended: handoff documents that survive restarts, a task router with a fallback backend, exponential back-off that refuses to retry auth errors, a sliding-window rate-limit guard, JSON-lines audit logs, stale-agent detection with a restart cooldown, and a two-step tmux notification protocol. Copy-ready Python and shell snippets plus a review prompt. No network calls or credentials included.
Un cycle Brainstorm → Spec validée → Plan à cases à cocher → TDD strict → Review de conformité, conçu pour les devs qui construisent des SaaS pilotés par IA (générateurs, agents, outils no-code) en HTML/CSS/JS vanilla ou Node/Express, avec intégrations multi-fournisseurs (Anthropic, OpenAI, Gemini, Mistral). Exemples d'usage : "ajoute une fonctionnalité X à mon générateur", "intègre un nouveau fournisseur IA", "refonds ce module"
Audit AI agent workspaces for authority, secrets, data flow, memory, supply-chain, and evidence risks—locally and without executing target code.
Check the YAML report a worker agent writes before a reviewer, dashboard, or scheduler trusts it. 13 checks with stable reason codes: parse errors, missing or placeholder fields, invalid status, completed_at without timezone / in the future / before assignment, 'did' longer than 3 lines, partial reports without target, actual and reason, free text outside block scalars, task_id mismatch. Human or JSON output, configurable contract, 14 tests. Read-only, local only.
Control Android and iOS devices from your AI agent using the Mobilerun framework for local and cloud orchestration.
Expert in end-to-end AI-assisted software delivery, from design engineering and design systems to shipping: design engineering, design system automation, technical documentation, spec-driven planning, polyglot code quality, systematic QA, release auditing, LLM context engineering, and Astro frontend optimization. Built for agent-agnostic workflows across the full 9-skill delivery lifecycle.
Download package that runs on your own computer inside your AI agent (e.g. Claude Code or Codex — its subscription/API costs are yours). Requires Python 3.8+ (standard library only). No required third-party services or fees; optional free form service and free hosting. Builds a one-file, mobile-ready landing page with SEO, structured data and a pre-launch check. 5 layouts, light/dark. No fake reviews. AI-written: check before publishing.