automation-audit-ops

Give an agent x402 payment execution with per-task budgets, spending controls, and a non-custodial wallet.

Route AI-agent products through architecture, harness design, orchestration, evaluation, recovery, operations, and bounded memory capabilities.

Audit the 12-layer agent stack for wrapper regression, memory pollution, tool-discipline failures and hidden repair loops when an LLM feature misbehaves.

Self-rate a completed non-trivial task on accuracy, completeness, clarity, actionability, and conciseness, with evidence per axis and a 1-5 scorecard.

Build a persistent multi-agent OS on a harness: kernel architecture, specialist agents, slash commands, file-based memory, scheduled automation, and state…

Turn a harness into an autonomous system with persistent memory, scheduled operations, computer use and task queuing on native crons and MCP.

Patterns for autonomous agent loops, from sequential pipelines to RFC-driven multi-agent DAGs.

REST/HTTP conventions: resource naming, method semantics, status codes, error envelopes, rate-limit headers, pagination, versioning.

Answer a question about the existing plan or build with citations, read-only: why did we, how does it work, what is built.

Diagnose a failed agent run reproducibly - capture, diagnosis, contained recovery, introspection report - instead of retrying.

Compare coding agents or configurations head-to-head on a YAML task suite in isolated worktrees: pass rate, cost, time, consistency.

Classify a workspace's skills, commands, agents, rules and hooks as daily (auto-load) or library (on demand) to keep the loaded set small.