ai-llm-app-attack

AI/LLM应用攻击:提示注入,Agent工具滥用RCE,RAG投毒,MCP供应链,torch.load pickle RCE。Use when testing LLM apps, agents, RAG, MCP plugins, or AI model file risks.

Quality

68%

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 ./skills/ai-llm-app-attack/SKILL.md

Quality

Content

43%Weight 40%Scale 1-5

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

The content is a dense brain-dump of the AI/LLM attack surface that is token-efficient but lacks structure, executable guidance, and a clear workflow with validation checkpoints. It reads as reference notes rather than an actionable skill body.

Suggestions

Break the single code block into section headers (e.g., 提示注入, Agent工具滥用, 模型文件) and move detailed vectors into a separate reference file with a one-level link from SKILL.md.

Add a short sequenced workflow with an explicit validate-before-reporting checkpoint (e.g., 1. discover endpoints, 2. probe vectors, 3. verify side effects via OOB, 4. record Fact only on verified impact).

Include at least one concrete, copy-paste-ready probe example (e.g., a sample prompt-injection payload or a torch.load unsafe-pickle snippet) to lift actionability.

Dimension	Reasoning	Score
Conciseness	The body is dense and token-efficient with no padding or basic-concept explanations, but the run-on pipe-separated lines inside a single code block trade clarity for brevity.	4 / 5
Actionability	It names concrete vectors and gives hints like '抓流量找/chat /agent /tool' and '实际触发工具副作用(OOB回连/读到文件)才写Fact', but provides no executable code or commands — mostly descriptive enumeration.	3 / 5
Workflow Clarity	There is no sequenced workflow or numbered steps, and the only validation hint ('验证:实际触发工具副作用') is buried mid-paragraph; per the destructive/attack-operation cap this cannot exceed a rough-sequence level.	2 / 5
Progressive Disclosure	The entire body is one monolithic code block with no section headers, no file references, and no navigation; no bundle files exist to provide deeper structure.	1 / 5
	Total	10 / 20 Passed

Description

92%Weight 40%Scale 1-5

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, complete, and distinct, clearly stating both the attack techniques covered and the triggering contexts in third person. It is slightly held back only by missing a few natural synonyms a user might say.

Dimension	Reasoning	Score
Specificity	Enumerates multiple concrete attack techniques — '提示注入', 'Agent工具滥用RCE', 'RAG投毒', 'MCP供应链', 'torch.load pickle RCE' — giving comprehensive coverage of the AI/LLM attack surface.	5 / 5
Completeness	Explicitly answers 'what' (the enumerated attack vectors) and 'when' (a clear 'Use when testing LLM apps, agents, RAG, MCP plugins, or AI model file risks' clause).	5 / 5
Trigger Term Quality	Includes natural trigger phrases ('testing LLM apps, agents, RAG, MCP plugins, or AI model file risks') but omits common synonyms like 'prompt injection', 'red team', or 'LLM security'.	4 / 5
Distinctiveness Conflict Risk	Targets a clearly distinct AI/LLM attack-surface niche with specific triggers that are unlikely to conflict with unrelated skills.	5 / 5
	Total	19 / 20 Passed

Validation

87%

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 — 14 / 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

	Total	14 / 16 Passed

Repository: Ed1s0nZ/CyberStrikeAI
Path: skills/ai-llm-app-attack/SKILL.md
Commit: 893f134

Reviewed: 2 days 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.