Drafts a single /goal line defining objectives, evidence, and stop conditions for coding agents.
npx skills add https://github.com/EpicenterHQ/epicenter --skill agent-goal
Please help me install this Agent Skill. Skill: agent-goal Source: https://github.com/EpicenterHQ/epicenter/tree/main/.agents/skills/agent-goal Command: npx skills add https://github.com/EpicenterHQ/epicenter --skill agent-goal
Long-running coding agents like Codex or Claude Code often drift without a clear objective, making it hard to know when work is actually done. This Skill produces a precise /goal line that gives the agent a concrete objective, observable validation evidence, and an explicit stop condition.
/goal line.bun test packages/auth exits 0 instead of vague claims./goal so the agent works in checkpoints, surfaces test output after each one, and stops only when the defined evidence appears.Ask the agent to write a /goal for migrating the auth package to the new API, with tests passing as the completion evidence.
High-intent search queries and answers about installing and using this skill.
Answer four questions in order: what should change, where the agent starts, what evidence proves it changed, and when it should stop. Then compress those answers into one line starting with /goal that names the lane, starting context, and observable completion evidence.
A good completion condition is judgeable from the transcript alone, using exact checks like a test command exiting 0 or a build succeeding. Vague statements like tests pass are rejected in favor of commands whose output the agent must surface.
Do not create a /goal for vague wishes, open-ended research, unrelated chores, or work with no observable evidence. The skill is scoped to slash-goal drafting only, not general prompt engineering, handoffs, or delegation briefs.
After three failed attempts on the same check, the agent should stop retrying, report the root cause, and present the next decision. This bounds repeated failure and keeps the transcript informative.
No. Long requirements should live in a file that the goal points to, rather than being pasted into the goal line. The goal itself carries only execution-critical detail: objective, lane, starting context, evidence, and stop condition.
HechaoYannetDraft a Quick Design Spec for small changes with acceptance criteria.
corvo007Run verification commands and validate exit codes before reporting completion.
QwenLMDrafts verifiable /goal objectives with binary done-when checks, guardrails, and budgets.
CodeAlive-AIWrite timestamped markdown entries for decisions, facts, insights, and completed tasks.