Implement YAML-based rules to control AI agent behavior.
npx skills add https://github.com/edictum-ai/edictum --skill edictum-oss-edictum-ai
Please help me install this Agent Skill. Skill: edictum-oss Source: https://github.com/edictum-ai/edictum/tree/main/.claude/skills/edictum-oss Command: npx skills add https://github.com/edictum-ai/edictum --skill edictum-oss-edictum-ai
This Skill enables precise control over AI agent behavior, ensuring that agents adhere to defined rules and boundaries, mitigating risks of unauthorized actions and data breaches.
Set up a rule to block access to sensitive files with the following YAML:
apiVersion: edictum/v1
kind: Ruleset
metadata:
name: block-sensitive-reads
defaults:
mode: enforce
rules:
- id: block-sensitive-reads
type: pre
tool: read_file
when:
args.path:
contains_any: [".env", ".secret", "credentials", ".pem", "id_rsa"]
then:
action: block
message: "Sensitive file '{args.path}' blocked."
High-intent search queries and answers about installing and using this skill.
Control AI agent behavior by implementing YAML-based rules that define deterministic boundaries. This enforces precise conditions on tool calls, mitigating risks of unauthorized actions and data breaches in production workflows.
YAML-based rule enforcement works by specifying rules that target specific tools and arguments, triggering actions like blocking access. It ensures deterministic decision-making and fail-closed architecture for production-grade security.
Yes, you can block an AI agent from reading sensitive files by creating a YAML ruleset. Configure a pre-rule to check if file paths contain sensitive keywords and apply a block action to prevent unauthorized access.
Yes, you need the Edictum runtime and server for full functionality. The architecture operates fail-closed with zero runtime dependencies, ensuring production-grade security and deterministic boundary enforcement.
The best way to define tool-call boundaries is using YAML-based rulesets with a fail-closed mode. This approach specifies exact tools and conditions, ensuring deterministic decision-making and mitigating risks in production AI workflows.
YAML rules provide deterministic, rule-based enforcement rather than dynamic contextual judgment. This means boundaries are strict and predefined, requiring explicit rule configuration for every condition an AI agent might encounter.
tanujsutariaAudit VitalArc views for design-token violations in read-only mode.
selfagencyAssess AI agent codebases against OWASP ASI Top 10 security risks.
zzw4257Runs static code analysis with customizable rulesets to detect security vulnerabilities across multiple languages.
velcraftingDefine and enforce authorization rules for protected actions with deny-by-default behavior.