How Hallmark's Pre-Emit Self-Critique Works: Six-Axis Quality Control Explained

Hallmark evaluates every generated artifact against six quality dimensions before emitting it, automatically revising any content that scores below 3 on any axis and stamping the final scores in a machine-readable comment.

The pre-emit self-critique is a quality gate built into the Nutlope/hallmark repository that ensures no artifact reaches the user without meeting minimum standards. According to the source definition in skills/hallmark/SKILL.md, every generated file undergoes systematic evaluation across six independent axes, with results persisted in a standardized stamp format documented in skills/hallmark/references/slop-test.md.

The Six-Axis Scoring System

Hallmark scores each artifact on a 1-to-5 scale across six dimensions:

• Philosophy (P) – Alignment with the brief’s conceptual intent
• Hierarchy (H) – Logical structuring and ordering of content
• Execution (E) – Technical correctness and completeness
• Specificity (S) – Level of detail and concrete guidance
• Restraint (R) – Conciseness and avoidance of unnecessary fluff
• Variety (V) – Diversity of ideas, formats, or visual treatments

Each axis is evaluated independently, allowing granular identification of weaknesses without conflating technical errors with conceptual misalignment.

Automatic Revision Logic

If any axis receives a score lower than 3, Hallmark automatically enters a revision pass. The system recomposes the artifact iteratively until all six scores meet the minimum threshold. This hard floor prevents low-quality output from reaching the user while allowing the system to self-correct specific deficiencies identified during the critique phase.

Machine-Readable Score Stamping

After passing the quality threshold, Hallmark records the six scores in a single-line comment at the very top of the file. The stamp follows the exact format specified in skills/hallmark/references/slop-test.md:

/* Hallmark · pre-emit critique: P5 H4 E5 S4 R5 V5 */

For example, a generated CSS file begins with:

/* Hallmark · pre-emit critique: P5 H5 E5 S5 R4 V5 */
body {
  margin: 0;
  font-family: system-ui, sans-serif;
}

The format uses single-digit integers (1-5) separated by spaces, prefixed with the axis letter, and enclosed in a comment block starting with /* Hallmark · pre-emit critique:.

Discovery on Future Runs

When Hallmark processes a brief that may regenerate existing artifacts, it scans for existing stamp comments to avoid repeating previous weaknesses. The system parses the stamp as a quick sanity check before generation, using historical scores to inform current output quality. This persistence mechanism ensures consistency across iterative development cycles.

Programmatically, the stamp can be validated using regular expression matching:

function shouldEmit(content) {
  const stamp = content.match(/\/\*\s*Hallmark\s·\spre-emit\scritique:\sP(\d)\sH(\d)\sE(\d)\sS(\d)\sR(\d)\sV(\d)\s*\*\//);
  if (!stamp) return false;               // no critique → reject
  const scores = stamp.slice(1).map(Number);
  return scores.every(s => s >= 3);        // all axes meet minimum
}

Summary

• Hallmark's pre-emit self-critique evaluates every artifact across six dimensions: Philosophy, Hierarchy, Execution, Specificity, Restraint, and Variety
• Scores below 3 on any axis trigger automatic revision passes until all thresholds are met
• Final scores are stamped in a machine-readable comment at the file's top, following the format defined in skills/hallmark/references/slop-test.md
• The system discovers and reads existing stamps on subsequent runs to prevent regression and maintain quality consistency

Frequently Asked Questions

What are the six axes in Hallmark's pre-emit self-critique?

The six axes are Philosophy (P) for conceptual alignment, Hierarchy (H) for logical structure, Execution (E) for technical correctness, Specificity (S) for detail level, Restraint (R) for conciseness, and Variety (V) for diversity of ideas. Each is scored 1-5 independently according to the definition in skills/hallmark/SKILL.md.

What happens if an artifact scores below 3 on any axis?

Hallmark automatically initiates a revision pass, recomposing the artifact until all six scores reach at least 3. This ensures no output reaches the user with critical deficiencies in any dimension, as implemented in the shouldEmit validation logic.

How does Hallmark read existing critique stamps from previous runs?

The system scans generated files for the specific comment pattern /* Hallmark · pre-emit critique: ... */ defined in skills/hallmark/references/slop-test.md. It parses these stamps to check historical scores and avoid repeating identified weaknesses in subsequent generations.

Where is the pre-emit self-critique defined in the Hallmark source code?

The process is defined in skills/hallmark/SKILL.md, which establishes the six-axis evaluation framework and revision logic. The exact stamp format is documented in skills/hallmark/references/slop-test.md, which specifies how Hallmark should read and write the critique comments.

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