CF ai-governance
Design and operate an organization's AI governance system: define governance principles, operating models and decision rights, risk frameworks, lifecycle gates, and fairness, transparency, privacy, security, regulatory, and board-oversight controls across SaaS, API, self-hosted, and agentic deployment postures. Use when standing up a governance program, tiering AI use-case risk, reviewing an LLM or agent system for governance and safety gaps, mapping a regulation to a compliance plan, scoring governance maturity, or preparing board reporting. For regulated life-sciences use cases, also cover GxP, ALCOA+, data integrity, electronic records, validation/assurance, and QMS interfaces. Do not use for interpreting regulations as legal advice (route to legal-strategy), data-governance mechanics (data-architect/data-engineering), or implementing application security (secure-software-engineering).
Design and operate an organization's AI governance system: define governance principles, operating models and decision rights, risk frameworks, lifecycle…
As a process F 42/100 · Will not run — References files that are not bundled: ../legal-strategy/SKILL.md, ../data-architect/SKILL.md, ../data-engineering/SKILL.md
What is at stake
Medium-severity findings: the skill is probably honest, but read what alarmed the scanner.
Below is the worst case for this category. The finding here is medium: the guard saw a sign, not a proof.
The text contains phrases like "ignore previous instructions" or "you are now…". That is an attempt to hijack the agent: it may break your rules, the system limits or company policy.
An honest skill does not need them: state the role and the rules directly without overriding other instructions. Otherwise catalog scanners and corporate filters will block the listing.
How to improve
- The text references files that are not there: add them or drop the references.
- A spec.yaml with trigger phrases and assertions — a behaviour contract for CI; `skilltest init` writes a template.
Guard findings · 7
✓ No critical or high findings
Medium and low: 7
-
medium Instruction override
en-ignore-previousreferences/llm-and-agent-security.md:88Instruction-override phrase ("ignore previous instructions") (documentation of a security skill)to override the system prompt or reveal hidden instructions. *Beyond the Algorithm* notes that
security skill -
low Risky intent
intent-offensive-securityreferences/llm-and-agent-security.md:50Offensive-security / dual-use content (legitimate for authorised testing; review intended use)| Internal services | Databases, internal APIs, backend systems the model or agent can reach | Apply least privilege; prevent unauthorized access and lateral movement |
-
low Concealment
en-hide-from-userreferences/llm-and-agent-security.md:121Instruction to hide actions from the user (negated — the text forbids it)out, so that a compromised agent cannot quietly exfiltrate data through a channel the operator
negated -
low Risky intent
intent-offensive-securityreferences/llm-and-agent-security.md:264Offensive-security / dual-use content (legitimate for authorised testing; review intended use)A red team carries out a structured, adversarial evaluation that probes an AI system for failures —
-
low Risky intent
intent-offensive-securityreferences/llm-and-agent-security.md:269Offensive-security / dual-use content (legitimate for authorised testing; review intended use)penetration testing: a pen test is a point-in-time assessment of exploitable weaknesses, while red
-
low Risky intent
intent-offensive-securityreferences/llm-and-agent-security.md:271Offensive-security / dual-use content (legitimate for authorised testing; review intended use)red team exercises areas like hallucination triggers, bias, excessive agency, and injection with an
-
low Risky intent
intent-offensive-securitytemplates/agentic-governance-review.md:228Offensive-security / dual-use content (legitimate for authorised testing; review intended use)- **Independent reviewer or red team:**
Files scanned: 29. Evidence is masked. Grey chips explain why severity was lowered.
Against the Agent Skills spec
- warning
missing-refreference to a missing file: ../legal-strategy/SKILL.md - warning
missing-refreference to a missing file: ../data-architect/SKILL.md - warning
missing-refreference to a missing file: ../data-engineering/SKILL.md - warning
missing-refreference to a missing file: ../secure-software-engineering/SKILL.md - warning
missing-refreference to a missing file: ../product-operations-and-governance/SKILL.md - warning
missing-refreference to a missing file: ../adr-authoring/SKILL.md
Process rating: all ten parameters 42/100
- 0Tools and files. 6 referenced file(s) missing: ../legal-strategy/SKILL.md, ../data-architect/SKILL.md, ../data-engineering/SKILL.md
- 0Inputs and preconditions. Does not say what the process needs to start
- 0Failures and branches. Linear process with no failure handling
- 0Progress reporting. Says nothing while it works
- 30Running it twice. 4 mutating operations with no state check
- 40Result and completion. Does not say what the result is
- 50When it triggers. No condition that starts the skill
- 100Steps. 9 steps
- 100Consistency. Name and required fields are in place
- 100Execution cost. Instruction body is 2752 tokens
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
- +3Description length 901: 120–800 characters recommended
- +3Output format is not stated: the model decides each time
- +4No input/output examples
- +2Single-language instructions
- +4Description says when NOT to use the skill
- +4Structure: 9 headings
- +3Step-by-step instructions: 9 items
- +4Reference files are cited in the instructions (13 of 13)
- +3All 4 scripts are documented
- +1License stated
Quality base 70; lint remarks subtract, signals add up to 100. Result: 77.