AgentSkillsCN

trace-parameter

从Streamlit的滑块/小部件一路追踪UI参数，直至其真正应用于计算的环节。找出那些虽已定义却从未被使用的孤立参数。

SKILL.md
--- frontmatter
name: trace-parameter
description: Trace a UI parameter from its Streamlit slider/widget all the way through to where it's actually used in computation. Finds orphaned parameters that are defined but never consumed.
argument-hint: "[parameter-name]"

Trace Parameter

Given a parameter name (provided as the argument, e.g. /trace-parameter bone_low), trace it through the entire codebase end-to-end. The argument is: $ARGUMENTS

Steps

1. Find UI Definition

Search app/main.py for where this parameter is defined as a Streamlit widget (slider, number_input, selectbox, checkbox, etc.). Record:

  • •The widget type and default value
  • •The st.session_state key if used
  • •The variable name it's assigned to
  • •File and line number

2. Find Where It's Passed

Search for where the variable is passed to function calls. Trace through:

  • •Direct function arguments in main.py
  • •Any intermediate wrapper functions
  • •Session state lookups that use it indirectly
  • •Record each handoff as file:line

3. Find Where It's Consumed

In the target core module (app/core/metal_detection.py, app/core/discrimination.py, app/contour_operations.py, etc.), find where the parameter is actually used in a computation (comparison, multiplication, array indexing, conditional, etc.). Not just received as a function argument — actually used.

4. Report the Chain

Format the trace as a chain:

code
PARAMETER: <name>

UI DEFINITION:
  app/main.py:<line> — <widget_type>(label="...", min=..., max=..., default=...)

PASSED THROUGH:
  app/main.py:<line> → function_name(<name>=variable)
  app/<module>.py:<line> → function_name(<name>=variable)

CONSUMED IN COMPUTATION:
  app/<module>.py:<line> — <how it's used, e.g. "mask = (hu_volume >= bone_low)">

STATUS: CONNECTED | BROKEN (parameter defined but never reaches computation)

5. If BROKEN

  • •Identify exactly where the chain breaks
  • •Show the function signature that should accept it but doesn't
  • •Suggest the minimal fix (add parameter to function signature, add it to the call site, add the computation)
  • •Check if there's a similar parameter that IS connected — the broken one may be a duplicate or renamed version

6. Check for Shadow Parameters

Look for other variables with similar names (e.g. bone_low vs bone_threshold_low vs bone_hu_min) that might cause confusion or indicate a rename that wasn't completed everywhere.

Important Notes

  • •If no argument is provided, scan ALL slider/number_input widgets in main.py and trace each one. Report a summary table.
  • •Pay special attention to parameters in the discrimination pipeline — these have known connectivity issues (see CLAUDE.md).
  • •Check both the "Russian Doll with Star Profile Discrimination" and "Russian Doll with Distance-Based Discrimination" code paths, as parameters may be connected in one but not the other.