用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt;运行前请先检查来源。
npx skills add https://github.com/Dev-Jahn/hippocampus --skill hippo命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地?可下载 SkillsMP 当前能够提供的文件。
Operating contract for parallel delegation waves — hand several tasks to external executors (codex exec) and subagents at once while main collects, verifies and merges. Use when the user says "launch in parallel", "fleet", "wave", "split it up", or "start everything you can". Worth reading for a single delegation too, when the call pattern or worktree isolation matters.
hippo project health check — inspects the ledger, the clerks, directive hygiene and waste patterns, then proposes cleanups. Use when the user asks for "/hippo:checkup", "check the project", "look for waste", "tidy the directives", or "how are the clerks doing".
基于 SOC 职业分类
正在显示 SKILL.md
| name | hippo |
| description | This is your hippocampus. Always use it. |
hippo does not control you. It just holds the memory for you: what you delegated, what was accepted or refuted, and which of the user's instructions are still alive. The judgment is always yours.
In Claude Code, hippo is on PATH. In Codex it is not — resolve ../../bin/hippo relative
to this SKILL.md into an absolute path and call that.
Facts go in through one door, hippo log <event>; the nouns are windows that read them back.
Call a noun on its own and you get its default view (hippo task = the list, hippo log = recent
records, hippo directive = the live instructions, hippo prior = the routing priors).
hippo task add <type>/<slug> --title "…"
(--deps a,b when order matters), hippo task done <id> when it ships. Bare hippo task
answers what can I start now — a task whose deps are unfinished shows a waiting on: line.hippo log dispatch --id <new id> --kind <tag> --exec <executor/model/effort> --scope "<one line>"
(a codex exec launched as hippo dispatch --kind … --scope … -- <codex args> records itself)hippo log outcome --ref <id> --result accepted|revised|refuted|no-go|lost --attr work|brief|harness
(--ref task:<task-id> resolves to that task's dispatch still awaiting an outcome, so you do not
have to go find the hash; the ledger still stores the dispatch id)hippo directive add --text "…" --lifetime turn|phase|durable
(add --audience main|executor|all when it binds only one side — a dispatched lane's capsule
carries the executor|all ones, this session's carries main|all)
(turn expires by itself after the next turn; phase and durable stay until
hippo directive withdraw <id>). To change an existing one, re-add it with the same --id —
a new id forks the instruction instead of updating it. Ids are lowercase kebab ascii
(gpu-pinning), whatever language the text is in; the derived id is refused when the text has
no ascii letters to build one from.hippo log review --id <new id> --base <sha> --source <where> --findings <n>
— and when its findings are dealt with, close the loop:
hippo log review-status --ref <review-id> --addressed full|partial|none [--at <sha>]
(a review with no review-status stays "not fully addressed" in every distill)hippo prior — which model and effort measured betterhippo statusDelegation briefs live in .hippo/briefs/ — a project-relative path that survives sessions,
so nothing has to retype an absolute scratchpad prefix. /hippo:dispatch has the full contract.
Every turn, a background clerk reads the transcript and infers most of the events above on its own. Recording through the CLI only raises the certainty. Nothing breaks if you skip a record — this organ does not enforce.
The clerk's backend resolves automatically (codex if installed, else headless claude). When a
project needs a different one, set clerk: {backend: codex|claude} in .hippo/config.yaml, or
export $HIPPO_CLERK_BACKEND / $HIPPO_CLERK_MODEL (pin the backend when you pin the model —
a model id for one backend is invalid on the other).