Transforms text through an iterative actor-critic loop until it cannot be distinguished from human writing and contains no logical fallacies. Uses a comprehensive 24-pattern detection catalog based on Wikipedia's "Signs of AI writing" guide.
This skill takes input text and runs an iterative actor-critic loop to transform it into natural, human-like writing while ensuring logical consistency.
The skill spawns two types of subagents in a loop:
The loop continues until the critic determines that:
INPUT TEXT → CRITIC → (needs work?) → ACTOR → CRITIC → ... → FINAL TEXT
↓ (pass)
OUTPUT
Receive the text to be humanized from the user. The text can be any length and on any topic.
Spawn the critic-agent subagent to evaluate the input text.
Use the Task tool to spawn the critic-agent:
critic-agent defined in agents/critic_agent.mdThe critic will return a structured evaluation containing:
verdict: Either PASS or NEEDS_REVISIONai_markers: List of detected AI-generated characteristics (if any)logical_issues: List of logical fallacies or inconsistencies (if any)confidence: How confident the critic is in their assessmentfeedback: Specific, actionable feedback for improvement (if verdict is NEEDS_REVISION)If the critic's verdict is PASS, proceed to Step 5 (Output).
If the critic's verdict is NEEDS_REVISION, proceed to Step 3.
Spawn the actor-agent subagent to rewrite the text.
Use the Task tool to spawn the actor-agent:
actor-agent defined in agents/actor_agent.mdThe actor will return the rewritten text.
IMPORTANT: The actor receives ONLY the critic's feedback, not the original evaluation details. This ensures the actor focuses on addressing specific issues rather than gaming detection criteria.
Spawn a NEW critic-agent subagent to evaluate the rewritten text.
This must be a fresh critic instance to ensure unbiased evaluation.
Return to Step 2.
Return the final text to the user along with:
To prevent infinite loops, the skill enforces:
The subagents are defined in the agents/ directory:
agents/critic_agent.md - The critic that evaluates textagents/actor_agent.md - The actor that rewrites textUser: /humanize
The implementation of machine learning algorithms necessitates a comprehensive
understanding of the underlying mathematical frameworks. It is important to note
that neural networks, in particular, require substantial computational resources.
Furthermore, the optimization of hyperparameters is a crucial step in the model
development process.
Expected behavior: