agentmods.dev

exp-down

A command for shutting down the AutoDL machine used by a research experiment. AutoDL is a service that provides rented computing instances.

In plain words
What is it for?
Use it to bring an experiment’s instance down after runs finish, or—only after confirmation—to kill in-flight runs and mark them accordingly.
Why use it?
It helps stop instance billing while first checking for running jobs and syncing their data back.

Command

1 19d ago A scan 18 tokens original MIT
View the source file zhyx12/projtool
Install

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.

agentmods
npx agentmods add commands/zhyx12/projtool/exp-down
Clone the repo
git clone --depth 1 https://github.com/zhyx12/projtool
Per session 18 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 453 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.00018 $0.00453
Opus 5 $0.00009 $0.00227
Sonnet 5 $0.00004 $0.00091
Haiku 4.5 $0.00002 $0.00045

Measured yesterday against content hash b27729454a96, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

exp-down 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 yesterday.

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.

src/projtool/assets/project_templates/claude/commands/exp-down.md · 44 lines

What it actually says

You are running the user's /exp-down command. The goal is to stop billing on the current experiment's instance without losing data.

Steps

  1. Identify the current experiment from the branch (exp/<topic> → exp-<topic>). If on main, ask the user which experiment to bring down.

  2. Call mcp__projtool__list_runs to confirm there are no status=running manifests on this experiment. If there are:

    • Tell the user which runs are still running.
    • Ask explicitly: "Take instance down anyway? In-flight runs will be killed and their manifests marked killed."
    • Do NOT default to yes.

  3. Flush the code and outputs sync sessions before tearing down — call mcp__projtool__sync_flush on:

    • the experiment's code session (<experiment_id>-code)
    • the global outputs session so any final artifacts are pulled back.

  4. Mark in-flight runs as killed via mcp__projtool__finalize_run with run_status="killed" for each (only if the user confirmed in step 2). The run_status arg name (not status) is required — status is reserved by the MCP boundary's ok/error wrapper.

  5. Call mcp__projtool__instance_down with experiment_id and the user's choice of:

    • power_off (default) — keeps the instance, just stops billing the GPU. Disk persists; next instance_up is fast.
    • release — fully releases the instance. Disk is gone. Use only when the experiment is conclusively done.

  6. Report what was done. Do not pause the mutagen session — projtool keeps it around so resumed instances pick up where they left off. If the user wants to fully terminate the sync, say so explicitly.

Tone

Short. The user just wants billing to stop.

Changes

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.

  1. yesterday First seen · 44 lines · 18 tokens per session scan A b27729454a96

Subscribe to this mod's changes

exp-down is a command published in the GitHub repository zhyx12/projtool (1 stars, last pushed 19d ago), licensed MIT. It adds 18 tokens to every session and 453 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.

Git and workflowResearch
Related

Other commands, from other repositories

checklist

Generate a custom checklist for the current feature based on user requirements.

github/spec-kit · 13 tokens

clarify

Identify underspecified areas in the current feature spec by asking up to 5 highly targeted clarification questions and encoding answers back into the spec.

github/spec-kit · 27 tokens

specify

Create or update the feature specification from a natural language feature description.

github/spec-kit · 13 tokens

analyze

Perform a non-destructive cross-artifact consistency and quality analysis across spec.md, plan.md, and tasks.md after task generation.

github/spec-kit · 26 tokens

constitution

Create or update the project constitution from interactive or provided principle inputs.

github/spec-kit · 13 tokens

converge

Assess the current codebase against the feature's spec, plan, and tasks, then append any remaining unbuilt work as new tasks to tasks.md so implement can complete it.

github/spec-kit · 35 tokens