moai-ref-llm-security

AI/LLM defensive security reference: prompt-injection defense, OWASP LLM Top 10 defensive mapping, MCP and agentic tool-call hardening, training-data poisoning detection, model-output validation and guardrails, MITRE ATLAS defensive correlation, and NIST AI RMF governance. Agent-extending skill that amplifies backend, security, and AI-application engineering with production-grade defensive patterns for LLM-backed systems. NOT for: offensive techniques (jailbreak authoring, attack-payload crafting, red-team exploitation), model training or fine-tuning methodology, prompt optimization for capability, web-app OWASP Top 10 (see moai-ref-owasp-checklist), or general API design (see moai-ref-api-patterns).

Quality

75%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Passed

No findings from the security scan

Fix and improve this skill with Tessl

tessl review fix ./.claude/skills/moai-ref-llm-security/SKILL.md

Quality

Content

50%

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

A well-organized, defense-only reference that avoids verbosity and clearly forbids offensive patterns, but it is a monolithic inline catalog without executable steps or progressive file-splitting. It sits at the mid-range across content dimensions.

Suggestions

Tighten or trim concept-explaining prose (e.g., the 'core defensive insight' paragraph and repeated untrusted-input framing) so every token earns its place.

Add a short, sequenced 'review workflow' with explicit validation checkpoints (inventory trust boundaries -> check OWASP LLM row-by-row -> run the Verification checklist) to lift workflow clarity toward a score of 3.

Move the longer tables (OWASP LLM Top 10, ATLAS correlation, NIST RMF) into reference files under references/ and keep SKILL.md as a concise overview with one-level-deep links to improve progressive disclosure.

Dimension	Reasoning	Score
Conciseness	Mostly efficient — dense tables pack many defensive controls into few tokens — but prose passages like 'The core defensive insight: an LLM does not distinguish data from instructions...' restate a concept Claude already knows, keeping it just short of fully lean.	2 / 3
Actionability	Guidance names concrete controls (instruction-hierarchy enforcement, provenance tagging, schema validation, ATLAS technique IDs) but is framed as reference knowledge rather than executable, copy-paste-ready steps, so it is actionable but not fully concrete per the score-3 anchor.	2 / 3
Workflow Clarity	There is a structured Verification checklist and Red Flags list, but the body is a reference catalog rather than a sequenced multi-step procedure with explicit validation checkpoints, so sequence/checkpoints are implicit rather than a clear ordered workflow.	2 / 3
Progressive Disclosure	Content is well-organized into labeled sections with cross-references to sibling skills, but everything lives inline in a single ~250-line SKILL.md with no bundle files and no one-level-deep reference split, so structure is present but not progressively disclosed.	2 / 3
	Total	8 / 12 Passed

Description

100%

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 third-person, concrete, and tightly scoped with explicit when-to-use triggers and clear out-of-scope boundaries. It scores at the top of the scale on every dimension.

Dimension	Reasoning	Score
Specificity	Lists multiple concrete defensive actions such as 'prompt-injection defense', 'OWASP LLM Top 10 defensive mapping', 'MCP and agentic tool-call hardening', 'training-data poisoning detection', and 'NIST AI RMF governance', matching the score-3 anchor for several specific concrete actions.	3 / 3
Completeness	The description states what the skill does, and the explicit when_to_use clause ('Use when hardening an LLM-backed application... designing guardrails... mapping a design against the OWASP LLM Top 10 or NIST AI RMF') clearly answers both what and when.	3 / 3
Trigger Term Quality	Covers natural terms a user would say — 'hardening an LLM-backed application or agent against prompt injection', 'designing guardrails or output validation', 'MCP/tool permissions', 'OWASP LLM Top 10', 'NIST AI RMF', and 'MITRE ATLAS' — giving good coverage of likely trigger phrasing.	3 / 3
Distinctiveness Conflict Risk	A clear LLM-defensive-security niche with explicit 'NOT for' carve-outs that point the user to moai-ref-owasp-checklist and moai-ref-api-patterns for adjacent domains, making mis-triggering unlikely.	3 / 3
	Total	12 / 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
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning

	Total	15 / 16 Passed

Repository: modu-ai/moai-adk
Path: .claude/skills/moai-ref-llm-security/SKILL.md
Commit: 05117b6

Reviewed: 1 day 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.