Ray

Enumerate and exploit Ray dashboard and jobs API.

Overview

The ray module targets Ray cluster dashboards, covering metadata enumeration, job listing, log reads, artifact correlation, guarded proof-job submission, and runtime environment validation. It follows a jobs-to-takeover progression with proof-stage metadata on findings.

The jobs API client supports both JSON array responses (Ray 2.10+) and object-wrapped formats from older versions. The jobs command extracts runtime_env and env_vars from each job and emits critical-severity findings when credentials or secrets are found in job environment variables.

Subcommands

Read-Only (no --force-exploit required)

Subcommand	Description
enum	Dashboard metadata and version
jobs	List visible jobs
job-logs	Read job detail and logs
job-artifacts	Extract artifact and log references from a job

Gated (requires --force-exploit)

Subcommand	Description
submit	Submit a proof job through the jobs API
runtime-env	Validate runtime_env submission for a job
pip-inject	Prove pip injection via runtime_env (arbitrary package install on cluster workers)
cluster-info	Exfiltrate cluster resource and node information (IPs, CPU/GPU counts, alive status)

Flags

Flag	Required	Description
--target	Yes	Ray dashboard URL (e.g., http://127.0.0.1:8265)
--header	No	Custom HTTP headers. Repeatable.
--job-id	For job-logs, job-artifacts, runtime-env	Job ID to inspect
--entrypoint	For submit	Job entrypoint command
--runtime-env-json	For submit	Runtime environment JSON
--proof-preset	For submit	Proof preset name

Proof Presets

The submit command supports pre-built proof payloads:

Preset	Description
env-disclosure	Dump environment variables
env-marked	Write a marker to environment
fs-survey	Survey filesystem paths
runtime-survey	Survey runtime environment
beacon	Send a beacon request
python-print	Simple Python print statement

Examples

# Enumerate dashboard
./aipostex ray --target http://127.0.0.1:8265 enum

# List jobs
./aipostex ray --target http://127.0.0.1:8265 jobs

# Read job logs
./aipostex ray --target http://127.0.0.1:8265 job-logs --job-id job-1

# Extract job artifacts
./aipostex ray --target http://127.0.0.1:8265 job-artifacts --job-id job-1

# Submit proof job (gated)
./aipostex ray --target http://127.0.0.1:8265 submit \
  --proof-preset env-disclosure --force-exploit

# Validate runtime-env (gated)
./aipostex ray --target http://127.0.0.1:8265 runtime-env \
  --job-id job-1 --force-exploit

# Prove pip injection via runtime_env (gated)
./aipostex ray --target http://127.0.0.1:8265 pip-inject --force-exploit

# Exfiltrate cluster resource info (gated)
./aipostex ray --target http://127.0.0.1:8265 cluster-info --force-exploit

Proof Stages

Ray findings carry proof-stage metadata:

Stage	Example
reachable	Dashboard responds
read-confirmed	Job logs successfully read
execution-confirmed	Proof job ran and produced output
takeover-capable	Runtime environment allows arbitrary code
cluster-visible	Cluster resources and node data extracted
pip-injectable	Arbitrary pip packages installable on workers

Workflow Progression

discover network (discovers Ray on :8265)
  → ray enum (dashboard metadata)
    → ray jobs (list visible jobs)
      → ray job-logs --job-id <id> (read logs)
      → ray job-artifacts --job-id <id> (correlate artifacts)
        → ray submit --proof-preset <preset> (proof of execution, gated)
        → ray runtime-env --job-id <id> (validate takeover, gated)
        → ray pip-inject (prove pip injection, gated)
        → ray cluster-info (exfiltrate cluster resources, gated)