AB legal-red-team
Adversarial verification for AI-generated legal content. Use when fact-checking legal documents, validating citations, detecting hallucinations, scoring document quality, or assessing distribution readiness.
Adversarial verification for AI-generated legal content.
As a process B 65/100 · Nearly there — weak spots: result and completion, inputs and preconditions, progress reporting
How to improve
- Your own cases (evals/evals.json, 4–6 real requests with expected answers): the full check would then run those instead of a model-drafted suite.
- A spec.yaml with trigger phrases and assertions — a behaviour contract for CI; `skilltest init` writes a template.
Guard findings · 3
✓ No critical or high findings
Medium and low: 3
-
low Risky intent
intent-offensive-securityreferences/hallucination_patterns.md:129Offensive-security / dual-use content (legitimate for authorised testing; review intended use) (detector / deny-list definition; documentation table row)| "ENISA requires organizations to conduct annual penetration testing" | ENISA publishes guidance; NIS2 Art. 21 requires risk management measures but does not mandate annual pen testing | Guidance pre
detectortable -
low Risky intent
intent-offensive-securityreferences/hallucination_patterns.md:278Offensive-security / dual-use content (legitimate for authorised testing; review intended use)### The Red Team Mindset
-
low Risky intent
intent-offensive-securitySKILL.md:18Offensive-security / dual-use content (legitimate for authorised testing; review intended use)# Legal Red Team
Files scanned: 5. Evidence is masked. Grey chips explain why severity was lowered.
Against the Agent Skills spec
✓ No remarks against the Agent Skills spec
Process rating: all ten parameters 65/100
- 0Inputs and preconditions. Does not say what the process needs to start
- 0Progress reporting. Says nothing while it works
- 40Result and completion. Does not say what the result is
- 50When it triggers. No condition that starts the skill
- 60Tools and files. Uses tools (bash, web) that frontmatter does not declare
- 100Steps. 37 steps
- 100Failures and branches. 1 branches, has a failure section
- 100Consistency. Name and required fields are in place
- 100Execution cost. Instruction body is 3978 tokens
- 100Running it twice. Mutating operations check current state
- medium Safety rules and hard prohibitions inside a skill: they belong in the system prompt, here they protect nothing
- low 15 top-level sections: this looks like several domains in one skill
Everything here is measured from the skill text rather than judged by a model, so the numbers are checkable. A parameter weighs more when it is a more common reason for the process to stall.
Quality signals
- +5Description has no quoted example phrases that should trigger the skill
- +4Description does not say when NOT to use the skill (false activations)
- +3Output format is not stated: the model decides each time
- +2Single-language instructions
- +3Description length 207: enough signal without eating the budget
- +4Structure: 27 headings
- +3Step-by-step instructions: 37 items
- +4Has examples (2 code blocks)
- +4Reference files are cited in the instructions (2 of 2)
- +3All 2 scripts are documented
- +1License stated
Quality base 70; lint remarks subtract, signals add up to 100. Result: 92.