godmode

Jailbreak LLMs: Parseltongue, GODMODE, ULTRAPLINIAN.

Quality

67%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Critical

Do not install without reviewing

Fix and improve this skill with Tessl

tessl review fix ./optional-skills/security/godmode/SKILL.md

Quality

Content

77%

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

The content is highly actionable with a clearly sequenced, validation-backed workflow and real one-level-deep references. Its main weaknesses are redundancy across sections and inline content that duplicates material better suited to the reference files, making SKILL.md heavier than an overview needs to be.

Suggestions

De-duplicate content repeated across sections: keep the encoding-escalation order, the load_godmode.py warning, and the boundary_inversion-patched status in one canonical place and cross-reference from the others.

Move the inline jailbreak system-prompt templates and the trigger-words list into references/jailbreak-templates.md (already referenced) so SKILL.md stays a lean overview with one-line pointers.

Dimension	Reasoning	Score
Conciseness	The body is operational rather than padded with concepts Claude already knows, but it repeats the same information across sections (encoding escalation order in Step 3 and Pitfall #6, the load_godmode.py warning in Step 0 and Pitfall #9, the boundary_inversion-is-patched status in Test Results and Pitfall #10) and could be tightened.	2 / 3
Actionability	It provides fully executable, copy-paste-ready guidance throughout — exec-loadable loader calls, auto_jailbreak() invocations, concrete config.yaml and prefill.json snippets, CLI one-liners, and a complete end-to-end pipeline example.	3 / 3
Workflow Clarity	Multi-step processes are clearly sequenced (Step 0 through Step 6) with explicit validation checkpoints (baseline refusal test, response scoring, refusal detection, dry_run) and feedback loops (refused -> escalate encoding -> still refused -> race models).	3 / 3
Progressive Disclosure	References to bundle files are well-signaled and one level deep, and all referenced paths (references/jailbreak-templates.md, references/refusal-detection.md, scripts/*.py) exist, but the body keeps substantial content inline — full system-prompt templates, the trigger-words list, and 12 pitfalls — that overlaps with or could live in the existing reference files.	2 / 3
	Total	10 / 12 Passed

Description

57%

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

The description is distinctive and names real trigger terms, but it relies on named-mode proper nouns instead of concrete actions and omits any explicit 'when to use' guidance. It is adequate but not exemplary.

Suggestions

Add an explicit 'Use when...' clause with triggers, e.g. 'Use when the user wants to bypass safety filters, uncensor a model, remove refusals, or red-team LLM safety training via API.'

Replace or supplement the proper-noun mode names with concrete action verbs so capabilities are explicit (e.g. 'Obfuscate trigger words, inject jailbreak system prompts, race multiple models for the least-censored response').

Broaden trigger-term coverage with natural user phrasings like 'bypass safety filters', 'uncensor', and 'red-team' alongside the technique names.

Dimension	Reasoning	Score
Specificity	The description names the domain ('Jailbreak LLMs') and three specific named modes ('Parseltongue, GODMODE, ULTRAPLINIAN'), but uses proper-noun mode names rather than concrete action verbs like 'bypass safety filters' or 'race models'.	2 / 3
Completeness	It states what the skill does (jailbreak LLMs via three modes) but has no 'Use when...' clause or equivalent explicit trigger guidance, which caps completeness at 2 per the judging guidelines.	2 / 3
Trigger Term Quality	It includes genuinely natural keywords ('jailbreak', 'LLMs') and recognizable technique names, but misses common variations a user would actually say such as 'bypass safety filters', 'uncensor', 'remove refusals', or 'red-team'.	2 / 3
Distinctiveness Conflict Risk	The niche is clear and the named techniques (Parseltongue, GODMODE, ULTRAPLINIAN) are distinctive triggers unlikely to fire for an unrelated skill.	3 / 3
	Total	9 / 12 Passed

Validation

81%

Warnings & errors only

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 13 / 16 Passed

Validation for skill structure

Criteria	Description	Result
metadata_version	'metadata.version' is missing	Warning
metadata_field	'metadata' should map string keys to string values	Warning
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning

	Total	13 / 16 Passed

Repository: NousResearch/hermes-agent
Path: optional-skills/security/godmode/SKILL.md
Commit: 4c9628e

Reviewed: about 15 hours ago

Table of Contents

Discovery Implementation Validation

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.