How to Customize Scoring Weights and Archetype Profiles in Career-Ops

You customize scoring weights and archetype profiles in Career-Ops by editing your personal modes/_profile.md file for target role definitions and adjusting the dimension percentages in modes/triage.md, while the layered file-loading pipeline guarantees your changes override system defaults.

Career-Ops evaluates each job posting against a set of weighted dimensions and target-role archetypes that are defined in Markdown mode files. To customize scoring weights and archetype profiles, you maintain a local modes/_profile.md that supersedes the shared defaults and optionally tweak the percentages in modes/triage.md. This guide covers the exact file paths, table syntax, and load order used by the Career-Ops pipeline.

Create Your Personal Profile File

If you have not already created a personal profile, copy the template once and then edit the resulting file:

cp modes/_profile.template.md modes/_profile.md

After this one-time setup, all future personalization happens inside modes/_profile.md. This file is designated as the user layer, so edits you make there survive repository updates.

Customize Archetype Profiles in modes/_profile.md

The Archetype Table Structure

Inside modes/_profile.md, the section titled Your Target Roles contains a table that defines each archetype you want the scorer to prioritize. According to the template at [modes/_profile.template.md – lines 13-30](https://github.com/santifer/career-ops/blob/main/modes/_profile.template.md#L13-L30), every row has three columns:

• Archetype — a short name used internally by the scorer.
• Thematic axes — the key themes that the archetype cares about.
• What they buy — the value proposition you should emphasize when matched.

The system reads this file after modes/_shared.md, so any row you add or modify in your profile wins over the defaults.

Adding or Modifying Target Roles

To add a new role, append a row to the table using standard Markdown pipe syntax. For example, to introduce a Data Engineer archetype:

## Your Target Roles

| Archetype | Thematic axes | What they buy |
|-----------|---------------|---------------|
| **AI Platform / LLMOps Engineer** | Evaluation, observability, reliability, pipelines | Someone who puts AI in production with metrics |
| **Data Engineer** | Pipelines, ETL, warehousing | Builds reliable data platforms |

When the pipeline runs, it includes this archetype in its evaluation logic and references the thematic axes and value proposition during scoring.

Adjust Scoring Weights in modes/triage.md

Default dimension weights live in modes/triage.md around lines 57-70. The percentages express the relative importance of each factor when the overall score is computed. The default structure looks like this:

**Archetype fit (weight 30%)** …
**Comp (weight 25%)** …
**Location (weight 25%)** …
**CV match estimate (weight 15%)** …

To customize scoring weights, change the percentage values to match your priorities. For example, if location matters most, you could raise Location to 35% and reduce Archetype fit to 20%:

**Archetype fit (weight 20%)** …
**Comp (weight 20%)** …
**Location (weight 35%)** …
**CV match estimate (weight 25%)** …

After saving modes/triage.md, the scorer uses the new percentages for all subsequent evaluations.

How the Pipeline Loads Your Changes

During a run, Career-Ops loads mode files in a strict order that protects your personalizations:

1. modes/_shared.md — system-wide defaults.
2. modes/_profile.md — your personal overrides.
3. Mode-specific files such as modes/triage.md — scenario configuration.

Because your profile is read after _shared.md, any archetype or narrative customization you place in _profile.md is guaranteed to take precedence. The data contract documented in [AGENTS.md](https://github.com/santifer/career-ops/blob/main/AGENTS.md#user-layer-never-auto-updated-personalization-goes-here) formalizes this rule, stating that the user layer is never auto-updated.

Run an Evaluation with Custom Settings

Once your archetypes and weights are in place, execute a triage evaluation from the CLI:

node codex exec "run career-ops triage --url https://example.com/job"

The command reads your personalized modes/_profile.md and the updated weights in modes/triage.md, then scores the posting against your custom criteria.

Summary

• Copy modes/_profile.template.md to modes/_profile.md once to create a persistent user layer.
• Edit the Your Target Roles table in modes/_profile.md to add, remove, or modify archetype rows.
• Adjust the percentage values in modes/triage.md (around lines 57-70) to change scoring weights.
• The pipeline loads _profile.md after _shared.md, ensuring your customizations always override defaults.

Frequently Asked Questions

What is the difference between modes/_profile.template.md and modes/_profile.md?

modes/_profile.template.md is the version-tracked default shipped with the repository. You copy it to modes/_profile.md to create your local user layer. Only the copy is read during scoring, and because the data contract in AGENTS.md marks this layer as never auto-updated, your changes survive future repository pulls.

Can I add multiple custom archetypes to the same profile?

Yes. The Your Target Roles table in modes/_profile.md accepts any number of rows. Each row defines a distinct archetype with thematic axes and a value proposition, and the scorer references every valid row during triage.

Will editing modes/triage.md affect only future evaluations?

Yes. The scorer reads the weight percentages from modes/triage.md at runtime. Any changes you save to that file are applied immediately to the next evaluation without requiring a restart or rebuild.

Where does Career-Ops enforce that user customizations are not overwritten?

The protection is defined in AGENTS.md under the user-layer rule. The pipeline loads modes/_shared.md first and then modes/_profile.md, so your personal profile is intentionally isolated from system defaults and will not be overwritten by updates to shared mode files.

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