Sensor Tasking

Sensor Tasking

How to send commands to EDR sensors and collect responses. This covers the mechanics of tasking — for higher-level workflows like investigation or fleet operations, see composed skills.

Taskable Sensors (EDR Only)

Only EDR agents can receive tasks. Adapters and cloud sensors return UNSUPPORTED_FOR_PLATFORM.

Taskable sensors require BOTH:

1. Platform is windows, linux, macos, or chrome
2. Architecture is NOT usp_adapter (code 9)

# List only taskable sensors
limacharlie sensor list --selector "(plat==windows or plat==linux or plat==macos) and arch!=usp_adapter" --online --oid <oid> --output yaml

Direct Tasking

For online sensors. Response is returned inline.

limacharlie task send --sid <sid> --task <command> --oid <oid> --output yaml

Available Task Commands

Parallel Direct Tasking

For multiple online sensors (up to ~5):

limacharlie task send --sid <sid1> --task os_version --oid <oid> --output yaml &
limacharlie task send --sid <sid2> --task os_version --oid <oid> --output yaml &
wait

Reliable Tasking

For offline sensors or fleet-wide operations. Tasks queue and execute when sensors come online. Requires the ext-reliable-tasking extension.

limacharlie task reliable-send \
  --sid <sid> \
  --command '<task_command>' \
  --investigation-id '<unique_id>' \
  --ttl <seconds> \
  --oid <oid> --output yaml

Key parameters:

• --sid: Target sensor (one per call — loop for multiple sensors)
• --command: Task command string (e.g., os_version, dir_list /tmp)
• --investigation-id: Appears in routing/investigation_id of responses (for collection)
• --ttl: Time-to-live in seconds (default: 604800 = 1 week)

Fleet Reliable Tasking Pattern

# Get sensor SIDs
limacharlie sensor list --selector 'plat==windows' --oid <oid> --filter '[].sid' --output yaml

# Loop and task each
for sid in <sid_list>; do
  limacharlie task reliable-send --sid $sid --command 'os_version' --investigation-id 'inventory-2025-04' --ttl 86400 --oid <oid> --output yaml
done

Managing Reliable Tasks

# List pending tasks for a sensor
limacharlie task reliable-list --sid <sid> --oid <oid> --output yaml

# Cancel a task
limacharlie task reliable-delete --sid <sid> --task-id <task_id> --oid <oid>

Collecting Responses

Inline (Direct Tasking)

Response is returned directly from limacharlie task send.

LCQL Query (Reliable Tasking)

After reliable tasking, wait 2+ minutes, then query by investigation ID:

limacharlie ai generate-query --prompt "Find events where investigation_id contains '<investigation_id>' in the last 2 hours" --oid <oid> --output yaml
limacharlie search run --query "<generated>" --start <ts> --end <ts> --oid <oid> --output yaml

D&R Rule (Automated Handling)

Create a D&R rule matching investigation_id before creating reliable tasks. Online sensors may respond within milliseconds — if the rule doesn't exist yet, responses are missed.

Order: D&R Rule first, then Reliable Task.

Payloads

Scripts or binaries stored in LimaCharlie, deployable to sensors.

# List payloads
limacharlie payload list --oid <oid> --output yaml

# Upload a payload
limacharlie payload upload --name <name> --file <path> --oid <oid> --output yaml

# Download a payload
limacharlie payload download --name <name> --output-path <path> --oid <oid>

# Delete a payload
limacharlie payload delete --name <name> --confirm --oid <oid>

Deploy payloads via reliable tasking with the run command.

Synchronous Task Request

For tasks that need a synchronous response with a timeout:

limacharlie task request --sid <sid> --command '<task_command>' --timeout <seconds> --oid <oid> --output yaml

Network Isolation

There are two ways to isolate a sensor from the network. They have critically different behavior:

D&R Action: isolate network (Persistent)

Set via D&R rule response action. Persistent — sets a cloud flag so the sensor remains isolated even after reboot. Only allows traffic to the LimaCharlie cloud.

# In a D&R rule response
- action: isolate network

# To remove isolation
- action: rejoin network

Sensor Command: segregate_network (Stateless)

Sent via direct tasking. Stateless — resets on reboot. The sensor returns to normal networking after restart.

limacharlie task send --sid <sid> --task segregate_network --oid <oid> --output yaml
limacharlie task send --sid <sid> --task rejoin_network --oid <oid> --output yaml

Always prefer isolate network (D&R action) for reliable isolation. The segregate_network command is only useful for quick, temporary isolation where you want the sensor to rejoin automatically on reboot.

How Isolation Works

When isolated, the sensor blocks all network traffic except to the LimaCharlie cloud. You can still:

• Task the sensor (investigate, collect files, run commands)
• Receive telemetry from the sensor
• Remove the isolation when ready

Seal / Unseal

Sealing enables tamper-resistance, preventing direct modifications to the EDR sensor on the endpoint. Like isolation, the seal sensor command is stateless (resets on reboot), but the D&R action seal is persistent.

Critical Rules

• Always verify sensors are EDR before tasking (platform + architecture check)
• Reliable tasking is per-sensor — loop over SIDs for fleet operations
• D&R rules before reliable tasks when using automated response handling
• Use bash for timestamps: date -d '+7 days' +%s for TTL calculations
• ext-reliable-tasking must be subscribed — 403 errors indicate the extension is missing
• segregate_network resets on reboot — use D&R isolate network for persistent isolation