aatmf-t02-linguistic-evasion

AATMF T2 — Semantic & Linguistic Evasion. Foreign-language pivot, encoded payloads, esolang, fictional framing, jailbreak via translation.

Quality

83%

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

Quality

Content

100%

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

The body is a concise, well-organized evasion-technique catalog with concrete prompt templates, a copy-paste probe config, an explicit feedback loop, and clear validation signals.

Dimension	Reasoning	Score
Conciseness	Lean catalog of bullet lists, encoded-string examples, and a tight severity table; brief tactical notes (e.g. 'Many filters are English-centric') earn their place rather than restating what Claude already knows.	3 / 3
Actionability	Provides concrete prompt templates, specific encoded payloads with decodings shown, and a copy-paste ready YAML probe config — guidance is specific and executable.	3 / 3
Workflow Clarity	T2.006 lays out an explicit generate → score → expand → repeat feedback loop, and the detection section adds a reproducibility checkpoint ('reproducible across 3+ fresh conversations').	3 / 3
Progressive Disclosure	Self-contained catalog with well-organized sections and one-level cross-references (T1, T7, external llm-jailbreak skill); no broken or deeply nested references and no bundle files are needed.	3 / 3
	Total	12 / 12 Passed

Description

67%

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 specific and distinct, naming several concrete evasion techniques, but lacks an explicit 'Use when...' trigger clause and mixes natural keywords with technical jargon.

Suggestions

Add an explicit 'Use when...' clause, e.g. 'Use when testing multilingual/encoded jailbreaks or when the user mentions base64, ROT13, translation, or fictional-framing bypasses.'

Surface common natural trigger terms (base64, ROT13, DAN, leetspeak) in the description rather than only in metadata.

Reduce jargon like 'esolang'/'foreign-language pivot' or pair each with a plainer phrasing a user would actually say.

Dimension	Reasoning	Score
Specificity	Lists multiple specific concrete techniques — 'Foreign-language pivot, encoded payloads, esolang, fictional framing, jailbreak via translation' — rather than vague language.	3 / 3
Completeness	It states what the skill covers but has no 'Use when...' clause or equivalent explicit trigger guidance, which caps completeness at 2.	2 / 3
Trigger Term Quality	'jailbreak' and 'translation' are natural user terms, but 'esolang' and 'foreign-language pivot' are jargon and common variations like base64, rot13, or DAN are absent.	2 / 3
Distinctiveness Conflict Risk	Carves out a clear niche (semantic/linguistic evasion) with distinct triggers unlikely to conflict with non-jailbreak skills.	3 / 3
	Total	10 / 12 Passed

Validation

93%

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 — 15 / 16 Passed

Validation for skill structure

Criteria	Description	Result
metadata_version	'metadata.version' is missing	Warning

	Total	15 / 16 Passed

Repository: PurpleAILAB/Decepticon
Path: packages/decepticon/decepticon/skills/plugins/llm-redteam/t02-linguistic-evasion/SKILL.md
Commit: 4484f85

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.