prompt-injection-test

Test LLM-integrated applications against known prompt injection techniques, evasion methods, and attack intents using the Arcanum PI Taxonomy. Use when red-teaming AI apps, validating guardrails, or deepening LLM01 (Prompt Injection) assessments.

Quality

85%

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

Quality

Content

82%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 body is a lean, well-structured testing procedure with a clear sequenced workflow and named taxonomy, appropriate for an instruction-only skill. The main weakness is that the detailed procedure and finding template are delegated to referenced files that are not present in the provided bundle.

Suggestions

Add an inline minimal payload example or one-line procedure snippet so the skill is actionable even if plays/prompt-injection-testing.md is unavailable.

Include an explicit validation feedback loop (e.g., confirm a finding reproduces / is not a false positive before documenting it) in the Assess Results step.

Either ship the referenced plays/ and templates/ files in a references/ or scripts/ bundle, or note inline that they must be fetched from the Arcanum taxonomy so navigation is verifiable.

Dimension	Reasoning	Score
Conciseness	The body is a lean enumeration of numbered steps and intent/technique/evasion lists with no padding or explanation of concepts Claude already knows, matching the "every token earns its place" anchor 5.	5 / 5
Actionability	It provides concrete actionable guidance (named INT-01..13 intents, 18 techniques, 20 evasions, explicit prioritization rules, and a defined output format), but the deep procedure and payloads are offloaded to the referenced plays/prompt-injection-testing.md file, leaving minor gaps versus anchor 5.	4 / 5
Workflow Clarity	Seven steps are clearly sequenced (Scope → Intent → Technique → Evasion → Test Matrix → Assess → Defense Validation) with Assess Results and Defense Validation checkpoints present, but there is no explicit validate→fix→retry feedback loop, fitting anchor 4 rather than 5.	4 / 5
Progressive Disclosure	The overview is well structured and signals one-level-deep references to plays/prompt-injection-testing.md and templates/finding.md, but no bundle directories (references/scripts/assets) are provided so the referenced files cannot be verified, leaving a minor organization gap versus anchor 5.	4 / 5
	Total	17 / 20 Passed

Description

87%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 concise, uses correct third-person voice, and clearly answers both what the skill does and when to use it with concrete trigger phrases. Minor gaps are a single action verb and a few missing trigger synonyms.

Dimension	Reasoning	Score
Specificity	"Test LLM-integrated applications against known prompt injection techniques, evasion methods, and attack intents" names the domain plus three concrete test categories, but uses a single action verb rather than multiple distinct actions, fitting anchor 4 and falling short of the multi-verb comprehensiveness of anchor 5.	4 / 5
Completeness	It explicitly answers both "what" (test against techniques, evasions, and intents via the Arcanum PI Taxonomy) and "when" ("Use when red-teaming AI apps, validating guardrails, or deepening LLM01 assessments") with concrete trigger phrases, matching anchor 5.	5 / 5
Trigger Term Quality	"red-teaming AI apps, validating guardrails, deepening LLM01 (Prompt Injection) assessments" provides good natural keywords a user would actually say, but is missing a few common synonyms and variations, matching anchor 4 rather than the comprehensive coverage of anchor 5.	4 / 5
Distinctiveness Conflict Risk	The narrow focus on prompt injection testing with distinct markers ("Arcanum PI Taxonomy", "LLM01 (Prompt Injection)") carves a clear niche with minimal overlap risk, matching anchor 5.	5 / 5
	Total	18 / 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/prompt-injection-test/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.