How to Convert False Agency Constructions to Active Voice Using Stop-Slop

The Stop-Slop skill eliminates false agency by applying rules from references/structures.md that rewrite inanimate subjects into active voice with human actors.

False agency constructions attribute human actions to abstract concepts or data points, creating vague and mechanical prose. The hardikpandya/stop-slop repository provides a declarative skill set that identifies these patterns and converts them into direct, active voice. By loading the rules defined in SKILL.md and its reference files, language models and writers can replace weak phrasing with clear authorial statements.

Understanding False Agency in Technical Writing

False agency occurs when inanimate objects or abstract concepts are assigned human verbs, implying action where none exists. This construction obscures responsibility and weakens clarity.

According to the source code in hardikpandya/stop-slop, these patterns are formally defined in references/structures.md under the False Agency section. The documentation identifies specific constructions that give human capabilities to non-human entities:

• "The data tells us" — Data cannot speak; humans interpret it.
• "A complaint becomes a fix" — Complaints do not act; teams resolve them.
• "The market rewards" — Markets are systems; buyers make decisions.
• "The conversation moves toward" — Discussions do not self-direct; speakers guide them.

Each entry includes a recommended "instead" formulation that names the human actor or switches to direct active-voice phrasing.

Implementation Methods

The Stop-Slop skill operates through textual rule definitions rather than code compilation. You can deploy these active voice conversions through three primary methods:

Upload as a Claude Skill

Add the entire repository folder to Claude as a custom skill. When loaded, Claude accesses SKILL.md and automatically references the linked references/structures.md file to detect and rewrite false agency patterns during conversation.

Embed in System Prompts

Copy the contents of SKILL.md directly into your system prompt. The model reads the core instructions, including the directive to consult the False Agency patterns, and applies the "instead" wording when generating or editing text.

Programmatic Application

For automated processing, read SKILL.md and the reference files via script, then feed the rules to any LLM that supports custom instruction sets. This method requires no repository changes; the transformation logic resides entirely in the rule definitions.

Practical Code Examples

Pattern-Based Rewriting in Python

The following implementation demonstrates how to programmatically convert false agency constructions using regex patterns derived from the Stop-Slop rules:

import re

def rewrite_false_agency(text: str) -> str:
    """
    Convert false agency patterns to active voice based on 
    stop-slop/references/structures.md rules.
    """
    patterns = {
        r"\bthe data tells us\b": "we read the data",
        r"\ba complaint becomes a fix\b": "the team fixed the complaint",
        r"\bthe market rewards\b": "buyers reward",
        r"\bthe conversation moves toward\b": "we steer the conversation toward"
    }
    
    for pattern, replacement in patterns.items():
        text = re.sub(pattern, replacement, text, flags=re.IGNORECASE)
    return text

# Example usage

input_text = "The data tells us that the trend is upward."
result = rewrite_false_agency(input_text)
print(result)  # Output: we read the data that the trend is upward.

System Prompt Configuration

When configuring a language model, include the specific directive from SKILL.md:

## False Agency

Giving inanimate things human verbs. **Instead:** Name the human. 
"The team fixed it that week" beats "the complaint becomes a fix."

With this instruction loaded, a model receiving "The data tells us that the trend is upward" will output "We read the data and see that the trend is upward."

Before-and-After Transformations

The references/structures.md file documents these canonical conversions:

• Before: "The decision emerges"
  After: "The team decided"

• Before: "The market rewards risk-taking"
  After: "Buyers reward risk-taking"

Key Repository Files

Understanding the file structure helps you locate specific rules and examples:

• SKILL.md — Core instruction set containing the False Agency directive and pointers to reference files.
• references/structures.md — Detailed pattern definitions, including the complete False Agency section with explanations and rewrites.
• references/phrases.md — Complementary list of banned phrases and adverbial crutches.
• references/examples.md — Before-and-after showcase demonstrating how the skill transforms various weak constructions.
• README.md — Repository overview and quick-start instructions.

Summary

• False agency constructions assign human verbs to inanimate objects or abstract concepts, creating vague prose.
• The Stop-Slop skill defines conversion rules in references/structures.md that replace fake agents with human actors.
• You can deploy these rules via Claude skill upload, system prompt embedding, or programmatic API integration.
• Pattern-based scripts using regular expressions can automate the conversion process offline.
• Active voice rewrites require naming the specific person or team responsible for the action.

Frequently Asked Questions

What is the primary source file for false agency rules?

The references/structures.md file contains the authoritative False Agency section that defines problematic constructions and their active-voice alternatives. This file is referenced by SKILL.md when the skill loads.

How does Stop-Slop convert phrases like "the data tells us"?

According to the rules in structures.md, the phrase "the data tells us" is rewritten as "we read the data" or "we interpret the data." This conversion removes the false attribution of speech to data and places the human interpreter as the grammatical subject.

Can I use Stop-Slop with language models other than Claude?

Yes. Because the skill is purely textual, you can embed the SKILL.md content into any system prompt or programmatically feed the rules to GPT-4, Llama, or other instruction-tuned models. The transformation requires no vendor-specific code implementations.

What distinguishes false agency from standard passive voice?

False agency uses active verbs but assigns them to inappropriate subjects (e.g., "the report highlights"), whereas passive voice obscures the subject entirely (e.g., "it was highlighted"). Stop-Slop addresses both by requiring explicit human actors: "the author highlights" or "we highlight."

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}