Experiment StatusSAFE

🔥 An autonomous AI agent that runs your deep learning experiments 24/7 while you sleep. Zero-cost monitoring, Leader-Worker architecture, constant-size memory.

The instruction file, verbatim from the audited commit — this is the text the model reads, and the surface the audit's instruction layer examines. Quoted here so you can judge it without cloning anything.

---
name: experiment-status
description: "Check status of running autonomous experiment loops"
---

# experiment-status

Check the current status of your autonomous experiment agent.

## Usage

```
Claude Code: /experiment-status
Claude Code: /experiment-status --project /path/to/project
Codex: $experiment-status
```

## Behavior

1. Read `PROJECT_BRIEF.md` — show the research goal
2. Read `MEMORY_LOG.md` — show key results and recent decisions  
3. Read `.cycle_counter` — show how many cycles completed
4. Check for running training processes via the configured execution backend
5. If training is running, tail the log file for latest output
6. Show GPU utilization through the configured backend
7. Check if `HUMAN_DIRECTIVE.md` exists (pending directive)

If `execution.mode=ssh`, controller state still comes from the local project
directory, but PID checks, training logs, and GPU status come from the
configured remote host.

## Output Format

```markdown
# Experiment Status — my-project

## Goal
Train ViT-B/16 on ImageNet to 78%+ accuracy

## Progress
- Cycles completed: 4
- Current best: 78.3% (Exp004, ViT-B/16 + cosine + mixup)
- Status: TRAINING (PID 12345, GPU 0, running 3.2h)

## Latest Training Log
Epoch 45/90 | loss: 2.134 | acc: 77.1% | lr: 1.2e-4

## Recent Decisions
1. [04-08 14:45] Target reached with mixup, trying stronger augmentation
2. [04-08 06:00] Cosine schedule helped, adding regularization

## Pending Directive
None (drop a file at workspace/HUMAN_DIRECTIVE.md to intervene)
```

SAFEgrade B · trust 89/100 Nothing in the source contradicts what it says it does. Grade A is reserved for packages that have also passed the behavioural sandbox.

What the source does

Filesystem

none-observed

Network

none-observed

Shell

none-observed

Dependencies

pinned

Secrets in source

none-found

Findings (0)

No findings outside the package's declared scope.

Gates applied: no_behavioural_pass.