Resolves ambiguity with evidence and executes tasks autonomously without asking clarifying questions.
npx skills add https://github.com/nsollazzo/ship-kit --skill yolo-nsollazzo
Please help me install this Agent Skill. Skill: yolo Source: https://github.com/nsollazzo/ship-kit/tree/main/.agents/skills/yolo Command: npx skills add https://github.com/nsollazzo/ship-kit --skill yolo-nsollazzo
--yolo flag) and DECISION mode for resolving a single delegated fork ("your call", "you decide"). - Resolve-don't-ask protocol: Classifies ambiguity as intent, approach, or fact, then answers it by reading code, running commands, checking git history, or searching the web instead of asking. - Multi-agent fan-out: Spawns parallel sub-agents (explorers, researchers, skeptics) for tasks with real breadth, then reconciles their outputs. - Safety floor: Still stops for confirmation on irreversible, outward-facing, or money-spending actions like force-pushes, unrequested PRs, or external messages. - Use Case: A user says "fix the flaky login test, yolo" — the agent investigates the failure, picks the most defensible fix, verifies it by running the test suite, and reports assumptions made, all without a single question. ## Quick Start Tell the agent "yolo, refactor the auth module and make the tests pass" to have it work autonomously to a verified end state.High-intent search queries and answers about installing and using this skill.
Invoke autonomous mode with a trigger phrase like "yolo", "just do it", or "figure it out". The agent then resolves every ambiguity by reading code, running commands, or researching docs, states its assumptions in one line, and proceeds to a verified end state.
FULL mode runs the entire task hands-off for the rest of the session, triggered by phrases like "yolo" or the --yolo flag. DECISION mode resolves only the decision currently on the table after phrases like "your call", without becoming a session-wide mode or spawning sub-agents.
Yes. A narrow safety floor remains: irreversible, outward-facing, or money-spending actions such as force-pushes, deleting production data, unrequested PRs, or sending external messages still require confirmation. Everything reversible is decided and executed directly.
Sub-agents are spawned only for tasks with real breadth, uncertainty, or multiple defensible approaches, such as audits, migrations, or broad reviews. Trivial work like typo fixes or single-file reads is handled in a single linear pass.
Yes. Specific standing constraints like "schema migrations require approval" or "never deploy on Friday" stay in force exactly as written. Only generic propose-versus-decide etiquette is overridden, and any conflict is surfaced rather than silently averaged.
PicoMLXCoordinate parallel read-only sub-agents to diagnose software bugs and rank root causes.
RediSearchCommissions an isolated sub-agent review of a pull request or commit before merging.
bravrosInvestigate defects with graphify and runtime probes to certify root causes without modifying code.
forbotsakeRun CEO, design, eng, and DX review skills sequentially with auto-decisions.