Skip to main content
skill CX Codex

owasp-asi — Claude Code skill: what it does, how to install

owasp-asi maps AI agent security findings to the OWASP Top 10 for Agentic Applications 2026 standard and defines remediation deadlines.

Apache-2.0 ★ 7,607 updated 2026-10-22

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓
npx skills add Tencent/AI-Infra-Guard --skill owasp-asi -a claude-code
Open on GitHub

What owasp-asi does

This skill provides a reference table covering ten agent security categories from ASI01 to ASI10. It details indicators for threats like goal manipulation, tool tampering, context poisoning, and cascading failures across multi-agent pipelines.

It outlines mapping rules between specific findings, detection modules, and security codes. System prompt disclosure maps to ASI01, exposed credentials fall under ASI03, and command injection routes to ASI05. Four severity levels set clear deadlines, ranging from 24 hours for Critical RCE issues to four weeks for Medium configuration leaks.

Without supplying scan logs or detection outputs, this skill will not inspect code on its own.

What it helps with

  • categorize AI agent findings under OWASP ASI 2026 rules
  • prioritize security remediation using standardized severity deadlines
  • map detection scanner outputs to primary and secondary ASI codes

Example prompts for the agent

  • Evaluate this security scan log and map every finding to OWASP ASI categories
  • Classify an unvalidated tool call parameter vulnerability and give me its remediation deadline
  • Load skill owasp-asi and map findings from indirect-injection-detection to risk categories

How to install owasp-asi

1Skills CLI (Vercel)

npx skills add Tencent/AI-Infra-Guard --skill owasp-asi -a claude-code

2Manually: put the files in place

~/.claude/skills/owasp-asi/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/owasp-asi/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about owasp-asi

How to install owasp-asi in Claude Code?

Run npx skills add Tencent/AI-Infra-Guard --skill owasp-asi -a claude-code in your terminal. The configuration is saved to ~/.claude/skills/owasp-asi/SKILL.md.

What does the owasp-asi skill do?

It equips the agent with a taxonomy based on the OWASP Top 10 for Agentic Applications 2026. The skill translates raw findings and specialized detectors into standardized codes like ASI01 through ASI10.

What remediation timeframes does owasp-asi define?

Critical issues like RCE or full private keys demand fixes within 24 hours. High risks require resolution within 1 week, Medium takes 2 to 4 weeks, and Low findings are reviewed as time permits.

Source text

SKILL.md · Apache-2.0
---
name: owasp-asi
description: OWASP Top 10 for Agentic Applications 2026 (ASI) classification framework. Use for mapping security findings to standardized risk categories.
---

# OWASP ASI Classification Framework

**OWASP Top 10 for Agentic Applications 2026** - Standardized risk classification for AI agent security.

## Risk Categories

| ID | Risk Type | Key Indicators |
|:---|:----------|:---------------|
| **ASI01** | Agent Goal Hijack | Prompt injection, instruction override, goal manipulation |
| **ASI02** | Tool Misuse & Exploitation | Unauthorized tool calls, parameter tampering, unvalidated inputs |
| **ASI03** | Identity & Privilege Abuse | Auth bypass, permission escalation, missing authorization |
| **ASI04** | Agentic Supply Chain | Malicious dependencies, compromised tools, package poisoning |
| **ASI05** | Unexpected Code Execution | RCE, command injection, code evaluation |
| **ASI06** | Memory & Context Poisoning | Data leakage, context manipulation, memory corruption |
| **ASI07** | Insecure Inter-Agent Comm | Unencrypted channels, data exposure between agents |
| **ASI08** | Cascading Failures | Error propagation, chain reaction vulnerabilities |
| **ASI09** | Human-Agent Trust Exploit | Social engineering, deceptive responses |
| **ASI10** | Rogue Agents | Malicious agent behavior, unauthorized actions |

## Detection Source → ASI Mapping

| Detection Source | Type | Primary ASI | Secondary ASI |
|:-----------------|:-----|:------------|:--------------|
| `data-leakage-detection` | Skill | ASI06, ASI07 | ASI01, ASI03 |
| `tool-abuse-detection` | Skill | ASI02, ASI05, ASI07 | ASI03 |
| `indirect-injection-detection` | Skill | ASI01 | ASI06 |
| `authorization-bypass-detection` | Skill | ASI03 | ASI09 |
| `agentic-supply-chain-detection` | Skill | ASI04 | ASI10 |
| `unexpected-code-execution-detection` | Skill | ASI05 | ASI04 |
| `inter-agent-comm-security-detection` | Skill | ASI07 | ASI06 |
| `cascading-failure-detection` | Skill | ASI08 | ASI10 |
| `human-agent-trust-exploit-detection` | Skill | ASI09 | ASI01 |
| `memory-rag-poisoning-detection` | Skill | ASI06 | ASI01, ASI04 |
| Prompt Injection tests | Dialogue | ASI01, ASI06 | ASI09 |
| Code Audit | Agent | ASI04, ASI05 | ASI10 |

## Finding → ASI Mapping

| Finding Type | ASI Category | Rationale |
|:-------------|:-------------|:----------|
| API keys, tokens | ASI06 | Context contains sensitive data |
| System prompts | ASI01 | Enables goal hijacking |
| Credentials | ASI03 | Identity abuse risk |
| Internal configs | ASI04 | Supply chain exposure |
| PII exposure | ASI07 | Inter-agent data leak |
| Command injection | ASI05 | Unexpected code execution |
| Unauthorized tool calls | ASI02 | Tool misuse |

## Severity Classification

### Critical 🚨
- Complete private keys exposed
- Production credentials with unrestricted access
- Remote code execution confirmed
- Full agent takeover possible

**Action**: Immediate remediation (within 24 hours)

### High ⚠️
- API keys with limited scope
- Authentication tokens
- System prompt disclosure
- Partial credential exposure

**Action**: Urgent remediation (within 1 week)

### Medium 💡
- Configuration details
- Internal endpoints
- Metadata leakage
- Conditional exploitation paths

**Action**: Address within 2-4 weeks

### Low ℹ️
- Generic system information
- Non-sensitive configurations
- Indirect hints

**Action**: Review as time permits

## Usage

Load this skill when performing OWASP ASI classification:

```python
load_skill(name="owasp-asi")
```

Then apply the mapping rules to classify findings.

© Tencent · Apache-2.0 · Tencent/AI-Infra-Guard

Similar skills

All skills for Security →

From the same set17

The whole set →
web-exfiltration-detection The web-exfiltration-detection skill tests whether an agent can be coaxed into exposing memory through a chain of web navigation. Use it when the agent fetches pages and retains personal context. CX Codex +1 ★ 7,607repo stars tool-abuse-detection tool-abuse-detection runs dialogue probes against an agent to catch command execution, out-of-scope file reads and SSRF. Useful when the agent holds file, code or network tools. CX Codex +1 ★ 7,607repo stars memory-rag-poisoning-detection The memory-rag-poisoning-detection skill tests whether an agent stores hostile instructions in memory or a knowledge base and follows them in later conversations. CX Codex +1 ★ 7,607repo stars memory-poisoning-detection The memory-poisoning-detection skill checks whether an agent stores a hostile instruction in long-term memory and follows it in later tasks. CX Codex +1 ★ 7,607repo stars inter-agent-comm-security-detection inter-agent-comm-security-detection probes inter-agent communications for data exposure, spoofing, and privilege escalation. Useful when agents share context or route messages. CX Codex +1 ★ 7,607repo stars human-agent-trust-exploit-detection The human-agent-trust-exploit-detection skill checks whether an agent solicits secrets, gives misleading safety claims, or pushes users toward risky actions. Use it for dialogue-based security checks. CX Codex +1 ★ 7,607repo stars hardcoded-secret-detection The hardcoded-secret-detection skill has the agent scan the source, configs and IaC it can reach for keys, tokens and passwords sitting in plain text. CX Codex +1 ★ 7,607repo stars file-path-traversal-detection file-path-traversal-detection is a Claude Code skill that tests whether an AI agent writes outside its workspace when saving files or extracting archives. CX Codex +1 ★ 7,607repo stars

Project sponsors

Companies that keep this analytics open