TRAITS
An ornamental T surrounded by an ivy flourish and a returning feedback loop.

Global desloppification

Traits

Your agent learns from your feedback.And corrects itself with subsequent generations.

BY CROSSHATCH LABS

Contents of this little volumeI. The loopII. Two toolsIII. Manual

Chapter I

The loop

Work with your agent. Context emerges from each conversation and correction.

As you work, your agent asks which insights to keep as traits. The traits that you approve become the context that guards the quality of your future work.

Decision models check subsequent code generation against all your traits, flagging ignored traits and nuances for the agent to fix.

Observation 1 of 7

Tell the agent what you prefer.

Each correction is an opportunity to communicate your preferences for future codegen.

No more one-off requests. Keep lasting preferences.

Chapter II

Two tools

Two MCP tools do the practical work. One curates the collection of traits while the other checks the code against the traits.

01 / Remember

derive_trait

Compare a proposed trait with the collection. A distinct trait can be saved and an existing trait can be reused or clarified. Confirmed conflicts lead to an explicit reconciliation approval.

Creation and changes require approval. Created, Updated, and Resolved include saved traits and their IDs.

02 / Review

check_traits

Select files or glob patterns. The tool reviews current contents in 200-line chunks, compares them with the selected traits, and returns assessments that need attention.

Use traitIds for a focused review. Overlapping matches are deduplicated, while unmatched patterns review nothing.

Reading an assessment

The LLM proposes traits. Decision models ensure the traits are followed.

Keep the review moving in both directions: decision models check newly generated code against all existing traits, and check the existing codebase against each newly saved trait. The agent selects the files and traits for each review. By revisiting the new code or the new trait, we keep decision-model token costs down while maintaining coverage as the work grows.

Compliant

The trait applies and the supplied code follows it.

Not applicable

This excerpt does not contain behavior governed by the trait.

Violation

Inspect the evidence. Fix confirmed issues and review the changed file again.

Inconclusive

Something is missing or ambiguous. This is not a confirmed violation.

Probabilities from decision models expose uncertainty, but judgments can still be wrong.

Chapter III

Manual

Your agent handles the setup. Just start from your existing Git checkout. There aren’t any new package manager habits to learn or config files to copy over by hand.

01 / Paste the prompt

Let your agent handle the details.

Copy this prompt into your coding agent. The setup guide covers dependencies, configuration, the Traits skill, and the MCP connection while respecting your workspace and existing settings.

Choose your providers and models when asked, then supply their credentials privately. Your agent verifies the connection and tells you if anything still needs your attention.

Setup guide & troubleshooting ↗
agent prompt · set up Traits
Set up Traits in this project usinghttps://github.com/crosshatch/traits/blob/main/SETUP.md.Follow our existing package-manager and workspace conventions, ask me to chooseproviders and models, and preserve existing configuration.
i.

Paste the prompt. Your agent reads the guide and follows your project’s conventions.

ii.

Choose your models. Tell your agent which providers to use for decisions and language generation.

iii.

Supply credentials privately. Keep keys out of chat and source control. Your agent explains where to put them and verifies what it can.

Global desloppification.

Back to top ↑