detrix-dynamic-observability

name	Detrix Dynamic Observability
description	This skill should be used when the user asks to "debug without print", "debug", "observe running code", "add metric", "inspect variables at runtime", "see what a variable is", or mentions "detrix", "logpoint", "dynamic metrics", "debugpy", "observe code". Debugger for agents.
license	MIT
metadata	{"version":"1.2.0","languages":["Python","Go","Rust"]}

Detrix: Observe Code Without Modifications

NEVER add print() or log.debug() statements. Use Detrix instead.

Detrix sets non-breaking logpoints via DAP protocol to capture values from running code without pausing execution or modifying source files.

Core Tools

observe - Quick Observation

Simplest way to add a metric. Auto-generates name, auto-finds line.

observe(file="auth.py", line=42, expression="user.id")
observe(file="auth.py", expression="user.id")  # auto-finds line if unique

Returns: Metric name (e.g., auth_py_42)

add_metric - Full-Featured Observation

Complete control with introspection options.

add_metric(
  name="debug_payment",
  location="@payment.py#250",
  expression="transaction.amount",
  connection_id="conn1",           # optional if single connection
  capture_stack_trace=true,        # capture call stack
  stack_trace_head=5,              # top N frames
  capture_memory_snapshot=true,    # capture local/global vars
  snapshot_scope="local",          # "local", "global", "both"
  mode="stream",                   # capture mode
  sample_rate=10,                  # for "sample" mode
  stack_trace_ttl=1800,            # auto-disable after N seconds
  group="auth"                     # organize metrics
)

enable_from_diff - Auto-Create from Git Diff

Parses print/log statements and creates metrics automatically.

enable_from_diff(diff="<git diff output>")

Recognizes: print(f"{x}"), fmt.Printf("%v", x), dbg!(x), console.log(x)

create_connection - Connect to Debugger

create_connection(host="127.0.0.1", port=5678, language="python")

Languages: python (debugpy), go (delve), rust (lldb-dap)

query_metrics - Get Captured Values

query_metrics(name="auth_py_42", limit=100)
query_metrics(group="auth", limit=50)

inspect_file - Find Correct Line

inspect_file(file_path="auth.py", find_variable="user")
inspect_file(file_path="auth.py", line=42)  # see available vars at line

All 29 Tools

Metrics: list_metrics, get_metric, update_metric, remove_metric, toggle_metric Groups: list_groups, enable_group, disable_group Events: query_system_events, acknowledge_events Connections: list_connections, get_connection, close_connection Diagnostics: validate_expression Config: get_config, update_config, validate_config, reload_config System: get_status, get_mcp_usage, disconnect_all Remote: wake, sleep

Capture Modes

Mode	Description	Use Case
stream	Every hit (default)	Debug single requests
sample	Every Nth hit	High-frequency loops
sample_interval	Every N seconds	Long-running processes
first	First hit only	Initial value capture
throttle	Max N/second	Rate limiting

Location Formats (All Tools)

All location-based tools accept these formats:

• file.py#42 - Simple format
• @file.py#42 - With @ prefix (optional)
• path/to/file.py#42 - With path
• file.py + line=42 - Separate parameter

Typical Workflow

# 1. Connect (if not auto-connected)
create_connection(port=5678)

# 2. Add observation
observe(file="auth.py", line=42, expression="user.id")

# 3. Wait for code to execute, then query
query_metrics(name="auth_py_42", limit=10)

# 4. Clean up when done
remove_metric(name="auth_py_42")

References

• Workflow Details - Debugger setup, capture modes
• Error Troubleshooting - Common errors and fixes