agent-security-audit

Audit AI agent configurations for security risks — excessive permissions, prompt injection surfaces, data exfiltration paths, and missing guardrails. Use when reviewing CLAUDE.md files, MCP configs, agent orchestration code, or any AI agent setup.

Quality

70%

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 ./plugins/ai-security-skills/skills/agent-security-audit/SKILL.md

Quality

Content

57%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 SKILL.md body is a concise, well-structured overview of a six-step audit methodology, but it delegates the substantive procedure and finding template to files that are absent from the bundle, undermining actionability, workflow detail, and navigation. Adding the missing plays/ and templates/ files (or inlining their key content) is the highest-leverage fix.

Suggestions

Provide the missing `plays/agent-security-audit.md` file (or inline the core procedure) so the delegated 'full procedure' is actually available.

Provide the missing `templates/finding.md` file (or inline a concrete finding-format example) so the Output section is actionable.

Add an explicit validation/checkpoint step to the workflow, e.g. confirming each finding against the actual agent configuration before producing the Recommendations.

Dimension	Reasoning	Score
Conciseness	The body is lean and assumes Claude's competence (it does not explain what prompt injection or OWASP LLM06 is), with one-line step descriptions and no padding; a few phrases in the OWASP reference list could be trimmed.	4 / 5
Actionability	Each step gives a specific checklist of what to examine ('Enumerate every tool, MCP server, file system path, network access, and credential'), but there is no executable code or commands and the promised 'full procedure in plays/agent-security-audit.md' references a file that is not present, leaving guidance incomplete.	3 / 5
Workflow Clarity	Six steps are clearly sequenced and each has a defined purpose, but there are no validation/checkpoint steps and the detailed procedure is delegated to a missing file, leaving the actual workflow detail inaccessible.	3 / 5
Progressive Disclosure	The overview is well-organized into Steps/Output/References with clearly signaled one-level-deep references, but both referenced files (plays/agent-security-audit.md, templates/finding.md) do not exist in the bundle, so navigation is effectively broken.	3 / 5
	Total	13 / 20 Passed

Description

83%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.

A strong description that answers both what and when with concrete, domain-specific trigger terms and enumerated risk categories. Its only weakness is mild breadth from the generic 'any AI agent setup' tail and a single action verb.

Dimension	Reasoning	Score
Specificity	Enumerates concrete audit targets ('excessive permissions, prompt injection surfaces, data exfiltration paths, and missing guardrails') but relies on a single action verb ('audit'), leaving minor coverage gaps relative to the comprehensive anchor.	4 / 5
Completeness	Clearly states what it does ('Audit AI agent configurations for security risks ...') and when to use it ('Use when reviewing CLAUDE.md files, MCP configs, agent orchestration code, or any AI agent setup'), matching the explicit what-and-when anchor.	5 / 5
Trigger Term Quality	Includes natural, concrete terms a user would say ('CLAUDE.md files, MCP configs, agent orchestration code'), giving good keyword coverage, though a few common synonyms are missing.	4 / 5
Distinctiveness Conflict Risk	Has a clear niche (AI agent security auditing) with distinct triggers, but the closing phrase 'or any AI agent setup' broadens scope enough to create minor overlap risk with related skills.	4 / 5
	Total	17 / 20 Passed

Validation

100%

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 — 16 / 16 Passed

Validation for skill structure

No warnings or errors.

Repository: OWASP/secure-agent-playbook
Path: plugins/ai-security-skills/skills/agent-security-audit/SKILL.md
Commit: 79fea6b

Reviewed: 4 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.