Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/dev-jahn/hippocampus/hipponpx skills add Dev-Jahn/hippocampus --skill hippogit clone --depth 1 https://github.com/Dev-Jahn/hippocampusWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/dev-jahn/hippocampus/hippo)<a href="https://agentmods.dev/skills/dev-jahn/hippocampus/hippo"><img src="https://agentmods.dev/badge/skills/dev-jahn/hippocampus/hippo.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00013 | $0.00932 |
| Opus 5 | $0.00006 | $0.00466 |
| Sonnet 5 | $0.00003 | $0.00186 |
| Haiku 4.5 | $0.00001 | $0.00093 |
Grade A, and why
hippo scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 61 lines — stays where its author put it; the contents beside it link to each section on GitHub.
hippo — the project's hippocampus
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.
How to call it
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.
The grammar, in one line
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).
When to reach for what
- When work lands that will outlive this turn:
hippo task add <type>/<slug> --title "…"(--deps a,bwhen order matters),hippo task done <id>when it ships. Barehippo taskanswers what can I start now — a task whose deps are unfinished shows awaiting on:line. - While delegating:
hippo log dispatch --id <new id> --kind <tag> --exec <executor/model/effort> --scope "<one line>"(a codex exec launched ashippo dispatch --kind … --scope … -- <codex args>records itself) - When a delegation gets a verdict:
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) - When the user gives a standing instruction:
hippo directive add --text "…" --lifetime turn|phase|durable(add--audience main|executor|allwhen it binds only one side — a dispatched lane's capsule carries theexecutor|allones, this session's carriesmain|all) (turnexpires by itself after the next turn;phaseanddurablestay untilhippo 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. - When an external review reply arrives:
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) - Before deciding delegation routing:
hippo prior— which model and effort measured better - To see where things stand:
hippo status
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 3d ago First seen · 61 lines · 13 tokens per session scan A ee57ac2b1c9e
hippo is a skill published in the GitHub repository Dev-Jahn/hippocampus (4 stars, last pushed 21d ago), licensed MIT. It adds 13 tokens to every session and 932 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
archon
Autonomous multi-session campaign agent. Decomposes large work into phases, delegates to sub-agents, reviews output, and maintains campaign state across context windows. Use for work that spans multiple sessions and needs persistent state, quality judgment, and strategic decomposition.
marshal
Meta-orchestrator that takes any direction — broad, specific, or vague — and autonomously chains skills and context into actionable work. Gathers context from codebase, docs, and memory. Only asks the user when it genuinely cannot proceed. Single-session orchestrator.
wiki
Markdown-first knowledge base where the LLM acts as librarian. Ingests raw sources, compiles and interlinks topic files, self-maintains an index. No vector DB or embeddings required -- uses LLM-native navigation over structured markdown up to 400K words.
memory-sync
Persist new context, terms, learnings, and settled lightweight decisions. Use when the user says remember this, save this for later, add to glossary, note this down, or at session end to consolidate what was learned — that goes to the memory layer (CLAUDE.md hot cache, docs/memory/, docs/inbox.md). Also use to record…
pocket-init
Onboards an existing (brownfield) project onto Pocket. Scans the codebase and writes a project memory file (CLAUDE.md or AGENTS.md), optionally enables Pocket Enterprise (mode init) and scaffolds GitHub issue/PR templates. Trigger on "pocket-init", "set up pocket", "onboard this project", "generate CLAUDE.md", "enable…
session-recap
Document Claude Code sessions by extracting knowledge into cross-referenced documentation. Triggers on "recap the session", "summarize the work", or after significant code changes.