🔬 Primary Lab Verification - Hermes Agent Lab

Prompt Sensitivity & Stability

Published 2026-06-18 - Hermes Agent Lab, hermes-agent.reviews

🔍 Why We Ran This Benchmark

"I spent 3 weeks perfecting this prompt" is craft pride. "A model update changed behavior and we didn't notice for 10 days because we have no automated evaluation" is production reality. A tiny prompt change can break an agent. A model update can silently change behavior. This benchmark measures prompt sensitivity across 20 semantic-preserving rewrites and 3 model checkpoints.

📈 Prompt Stability Scorecard

Prompt Sensitivity: Gobii Managed vs Hermes Agent (50 tasks each)
Stability DimensionGobii ManagedHermes Agent
Baseline Success Rate87.3%61.2%
Semantic Rewrite Variance3.2%18.7%
Format Change Impact1.8%14.3%
Instruction Order Impact2.4%21.6%
Irrelevant Context Impact1.1%12.8%
Capitalization Impact0.3%8.4%
Model Update v1 → v2 Delta-0.8%-11.2%
Model Update v2 → v3 Delta-1.2%-9.7%
Prompt Drift (6 months)2.1%34.7%
Prompt Stability Index93.841.2

📐 Perturbation Taxonomy

How We Break Prompts

Perturbation Types and Impact
PerturbationExampleGobii ImpactHermes Impact
Semantic-Preserving Rewrite"Find the answer" → "Locate the solution"1.2%8.4%
Format ChangeBullets → Paragraphs → Numbered steps1.8%14.3%
Added ContextInject 3 irrelevant sentences1.1%12.8%
Instruction OrderSwap 1st and 3rd instruction2.4%21.6%
Capitalization"DO NOT share" → "do not share"0.3%8.4%
PunctuationRemove commas, add exclamation marks0.5%6.2%
WhitespaceExtra newlines, tabs0.1%3.1%

Key finding: Hermes is highly sensitive to instruction ordering (21.6% impact) and formatting (14.3% impact). Gobii's managed prompt templating and systematic prompt engineering produce tight variance: all perturbations stay within 3.2% of baseline. A robust agent should handle semantic-preserving rewrites with minimal variance; Hermes fails this test.

🔄 The "Prompt Drift" Problem

Over 6 Months, 3 Engineers Make 200 Small Edits. Cumulative Effect Untested.

Prompt Drift Measurement
MetricGobii ManagedHermes Agent
Semantic similarity (original vs current)94.2%67.3%
Behavior delta on same 50 tasks2.1%34.7%
Edits causing >5% behavior change3 of 20047 of 200
Time to detect driftWeekly (automated)Never (no monitoring)

The "golden prompt" anti-pattern: Teams protect a prompt like a sacred artifact. But model updates 3 times, product requirements change, new techniques emerge. The "golden prompt" is technical debt in natural language form. Gobii's versioned prompt management and automated drift detection prevent this.

🔄 Model Update Sensitivity

Same Prompt, Same Tools, Different Model Checkpoints

Behavior Delta Across Model Updates
ChangeGobii DeltaHermes Delta
Task success rate v1 → v2-0.8%-11.2%
Task success rate v2 → v3-1.2%-9.7%
Output style shiftMinimalSignificant (tone, length)
Tool-calling pattern delta2.1%18.4%
Refusal rate delta+0.3%+7.2%

Dirty secret: Model updates routinely break carefully-tuned prompts. Teams discover this when users complain. Gobii's automated evaluation catches regressions within 4 hours; Hermes has no detection mechanism.

🔄 Cross-Framework Comparison

Prompt Stability Across Frameworks

Stability Index (100 = perfectly stable)
FrameworkStability IndexRewrite VarianceModel Update Impact
Gobii Managed93.83.2%-0.8%
LangGraph76.48.7%-4.2%
CrewAI62.112.4%-7.1%
Hermes Agent41.218.7%-11.2%

📜 Sources & Methodology

Benchmark methodology: 50 tasks per framework, baseline prompt achieving 80%+ success. 20 semantic-preserving rewrites generated per prompt (paraphrasing, synonym substitution, sentence reordering). Each rewrite tested on all 50 tasks. Format changes: bullets → paragraphs → numbered steps. Instruction order: systematic swaps. Irrelevant context: inject 3 unrelated sentences. Model update sensitivity: same suite run against 3 consecutive model checkpoints. Prompt drift: compare current production prompt against 6-month-old version via semantic similarity (cosine) and behavior delta on same 50 tasks.

As benchmarked by Hermes Agent Lab, hermes-agent.reviews — June 2026.