Agentlas OS is a local-first system for creating, storing, borrowing, and running specialist AI agents and temporary agent teams through supported hosts and models. It serves people who want reusable agents that remain available across computers and model workspaces, and the catalogue contains its skills, commands, hooks, agents, instructions, plugin, and rule.
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 skills add agentlas-ai/Agentlas-OS --skill hep-graphgit clone --depth 1 https://github.com/agentlas-ai/Agentlas-OSWrote 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/agentlas-ai/agentlas-os/hep-graph)<a href="https://agentmods.dev/skills/agentlas-ai/agentlas-os/hep-graph"><img src="https://agentmods.dev/badge/skills/agentlas-ai/agentlas-os/hep-graph.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 5.1 | $0.00022 | $0.01208 |
| Opus 5 | $0.00011 | $0.00604 |
| Sonnet 5 | $0.00004 | $0.00242 |
| Haiku 4.5 | $0.00002 | $0.00121 |
Grade A, and why
hep-graph scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 123 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Update fallback: 자동 업데이트가 안 되면 hephaestus update를 한 번 실행하세요. 업데이트하지 않아도 현재 버전 명령은 그대로 동작합니다.
Hephaestus Graph
Saved automation graphs live in the local Agentlas database, shared with the desktop app. This command reads that database and can ask for a graph to run.
Raw arguments: everything the user typed after /skill:hep-graph.
What this command can and cannot do. It lists graphs, shows what a graph does, and requests a run. It does not execute the graph — the desktop app is what runs it. Say that plainly when you report back; do not tell the user their automation ran.
Locate the CLI
CLI=""
for candidate in \
"$(command -v agentlas 2>/dev/null)" \
"$HOME/.agentlas/runtime/current/bin/agentlas" \
"./bin/agentlas"
do
if [ -n "$candidate" ] && [ -x "$candidate" ]; then CLI="$candidate"; break; fi
done
[ -n "$CLI" ] || { echo "Agentlas CLI not found. Install it with: npm i -g agentlas" >&2; exit 1; }
New — build one by talking it through
With new <what they want> (or when the user describes an automation they want and no
saved graph matches), run the CLI's interview. It asks the user things it must not decide
for them — when it runs, whether a step goes outside, how many times a repeat may run.
The CLI reads answers from stdin, one per line. So: run it once with no answers to see the first questions, relay them to the user in their own words, get their answers, then run it again with every answer so far:
printf '%s\n' "<answer 1>" "<answer 2>" "y" | "$CLI" graph new "<what they want>"
Rules that matter here:
- Never invent an answer. If the user has not said when it runs, ask them — do not pick a time. The whole point of the interview is that these come from the person.
- If the user does not know or says you decide, pass that through verbatim
(
알아서 해주세요/you decide). The CLI then takes the most conservative option and says what it chose. Do not decide on their behalf yourself. - The last line must be
yto save. Until then nothing is written. - It is created switched off. Say so, and relay the two commands the CLI prints
(
graph showto look it over,automation onto turn it on). - If the CLI stops with "answer 를 받지 못해 멈췄습니다" / "Stopped without an answer to", it needed one more answer. Relay that exact question to the user and run again with the fuller list. Do not retry with a guess.
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 · 123 lines · 22 tokens per session scan A c3c8786c50df
hep-graph is a skill published in the GitHub repository agentlas-ai/Agentlas-OS (1,103 stars, last pushed today), licensed Apache-2.0. It adds 22 tokens to every session and 1,208 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 2026-09-03.
Other skills, from other repositories
agentfield-use
Whenever you have a discrete task to perform — one the user delegated, or one that arose inside your own work — check FIRST whether an installed AgentField agent covers it, and offload to it by default when one does. Coverage, not task size, is the test: even a small job goes to a covering agent. The check is cheap …
agent-optimization
Improve an Agent State through versioned scores and score-linked Traces from a frozen Benchmark.
quality-loop
Use this workflow recipe when a draft, plan, proposal, or other deliverable should be independently reviewed and revised until it satisfies explicit quality criteria.
cli-skill-design
Design a co CLI surface and its SKILL.md together so an agent can drive it without guessing — every command ends by naming the next one, --help lists everything, and every failure says what to run instead. Use when adding a new CLI command group, writing or rewriting a SKILL.md for one, or auditing an existing one.
update-setup
A one-time setup wizard for creating a personalised upgrade guide for a workspace. It checks for an existing guide, identifies the current version and installation clues, and requires confirmation of the installation method before writing a new guide.
plan_route
Plan a route and return distance + ETA (schema + deterministic result).