Evenfire is an intelligence layer platform for running LLM agents that take real actions, on your infrastructure
TypeScript
/**
* gfs access review (spec plan §Security and compliance; plan P5-S03). Produces
* the periodic "effective grants per subject" report — for each subject, every
* grant it holds (with the resource and the inherit flag, which is what makes a
* grant reach a subtree). Pure aggregation over the grant rows.
*/
export interface GrantRecord {
subjectKey: string
resourceId: string
permissions: string[]
inherit: boolean
}
export interface SubjectGrant {
resourceId: string
permissions: string[]
/** Inheriting grants reach the whole subtree below resourceId. */
inherit: boolean
}
export interface SubjectAccess {
subjectKey: string
grants: SubjectGrant[]
/** Union of all permission bits the subject holds anywhere (review summary). */
effectiveBits: string[]
}
/**
* Group grants by subject into the access-review report. Subjects are returned
* in sorted order for stable reporting; each subject's grants list is the basis
* for "who can reach what", including inheriting (subtree-reaching) grants.
*/
export function effectiveGrantsPerSubject(grants: GrantRecord[]): SubjectAccess[] {
const bySubject = new Map<string, SubjectGrant[]>()
for (const g of grants) {
const list = bySubject.get(g.subjectKey) ?? []
list.push({ resourceId: g.resourceId, permissions: [...g.permissions], inherit: g.inherit })
bySubject.set(g.subjectKey, list)
}
return [...bySubject.entries()]
.sort(([a], [b]) => (a < b ? -1 : a > b ? 1 : 0))
.map(([subjectKey, subjectGrants]) => ({
subjectKey,
grants: subjectGrants,
effectiveBits: Array.from(new Set(subjectGrants.flatMap(sg => sg.permissions))).sort(),
}))
}