Skip to main content
AI/MLathola

latent-space-engineering

Shapes agent behavior via instruction framing and style transfer. Use when composing dispatch prompts or writing skill instructions for parallel review agents.

Stars
294
Source
athola/claude-night-market
Updated
2026-05-30
Slug
athola--claude-night-market--latent-space-engineering
View on GitHubRaw SKILL.md

// install — copy + paste into any project

mkdir -p .claude/skills && curl -fsSL https://raw.githubusercontent.com/athola/claude-night-market/HEAD/plugins/imbue/skills/latent-space-engineering/SKILL.md -o .claude/skills/latent-space-engineering.md

Drops the SKILL.md into .claude/skills/latent-space-engineering.md. Works with Claude Code, Cursor, and any agent that loads SKILL.md files from .claude/skills/.

Latent Space Engineering

Shape agent behavior by framing instructions for optimal performance. Distinct from context engineering (packing the right information), this skill addresses HOW instructions are framed to put agents in productive mental states.

When To Use

  • Composing agent dispatch prompts
  • Writing skill instructions that guide behavior
  • Dispatching 3+ parallel review agents
  • Generating code or documentation that must match an existing style

When NOT To Use

  • Packing factual context (use context-optimization)
  • Simple single-shot tasks with no behavioral nuance
  • Tasks where instruction tone is irrelevant

Core Techniques

1. Emotional Framing

Replace threat-based prompting with calm, confident instructions. Fear-based prompts cause rushing and corner-cutting.

Load module: modules/emotional-framing.md

2. Style Gene Transfer

Inject exemplar code or prose into context before requesting output. Agents reproduce stylistic attributes from pre-loaded samples.

Load module: modules/style-gene-transfer.md

3. Competitive Review

Frame multi-agent review dispatch with competitive incentives to increase rigor and thoroughness.

Load module: modules/competitive-review.md

4. Named Register Invocation

Name a published standard to pull a whole writing register into output at a cost of one line. Cheaper than exemplar injection and coarser, with a known mid-session decay.

Load module: modules/named-register-invocation.md

Quick Reference

Technique When Module
Emotional framing Any agent prompt emotional-framing
Style gene transfer Code/doc generation style-gene-transfer
Competitive review 3+ parallel reviewers competitive-review
Named register Operator or procedural text named-register-invocation

Writing for the Delegate

Framing shapes a delegate's output as much as it shapes a model's prose. Before dispatching, establish what model the delegate runs, what tools it has, and what project context it lacks, then write the prompt as a proposal rather than a procedure.

See modules/know-your-delegate.md.

Exit Criteria

  • When a named register is invoked (for example ASD-STE100), a deterministic check verifies the register held, per modules/named-register-invocation.md
  • Dispatch prompts contain no threat-based language ("you must", "don't fail", "or else"); replaced with calm, confident framing per modules/emotional-framing.md
  • When generating code or docs to match an existing style, at least one exemplar sample is injected into context before the output is requested (style gene transfer applied)
  • When 3+ parallel review agents are dispatched, each agent prompt includes a competitive framing element per modules/competitive-review.md
  • The correct technique module is loaded for the task type; irrelevant modules are not loaded (token efficiency maintained)