A kickoff protocol (agent skill) for multi-step tasks. Before writing any code, the agent must: run a spec interview, batch every human-authorization step into one checklist, build a self-verification loop (log sink + verify script + dry-run) *before* the feature itself, and build comparable variants instead of blocking on third-party decisions.
This Skill is built on practical expertise in AI agent safety, prompt-injection defense, source-to-sink privacy review, tool and MCP metadata integrity, memory provenance checks, autonomous action scoping, and evidence-based verification. It helps agents separate instructions from evidence before tool calls, uploads, publishing, deletion, code execution, or network output.
Prevents AI agents from declaring tasks complete until every requirement is verified with a one-line evidence anchor — when a gate fails, the agent stops and reports rather than proceeding, and only a human can unlock the gate to continue.
From a local ISSUE file to a verified, committed deliverable — with a fresh-subagent acceptance gate that catches wrong screenshots and leaked logic.
Your agent reads user-pasted text, web pages, retrieved docs, or tool output — and you can't tell which sentence an attacker wrote. A one-line "ignore malicious instructions" never held. This delivers a layered injection-defense design: threat model, attack-surface map, input/privilege/output defenses, canary tokens and a classifier, an adversarial red-team set with before/after, and residual risk plus an incident runbook.
Connect Cursor, Claude Code, Hermes & Codex with a shared local SQLite brain. Slash token costs by 92.2% with instant, zero-amnesia cross-agent recall. 100% private, BYOK, lifetime buyout.
I've spent months building my AI agent infrastructure with cost-efficient models (DeepSeek + Qwen), using them daily for content creation, strategy, and automation. I built SplitBrain after hitting the same wall every solo brainstorm: by round 3, a single AI just parrots you. Two complementary models, 10 minutes max, signal over noise. No corporate team. No committee. Just a self-taught creator who optimized the brainstorming loop and wants you to stop burning tokens on diminishing returns.
A lightweight Copilot subscription agent runtime built on FastAPI and Copilot SDK. Manage multi-account OAuth, scheduled tasks, GitHub Enterprise support, and a bilingual Web UI.
Turn messy notes, prompts, chat history, SOPs, and project docs into a reusable AI agent skill. Get a clean SKILL.md, reference-file plan, privacy review, and test prompts for Claude, Codex, Capafy, or local agents.