IAM Deceptive Escalation Auditor

Audit the union of every IAM policy attached to one principal for privilege-escalation paths that no single statement reveals, and for apparent escalations that are already neutralised. Resolves the effective permission set across all attached policies (Allow minus blanket Deny), then checks the cross-statement escalation combos (iam:PassRole + a compute-launch action, policy-rewrite-in-place, function-code hijack, self-attach admin, trust-policy rewrite + assume, credential minting for another identity), the wildcard grants (Action '*' on Resource '*', service-level wildcards, Allow+NotAction), and the trust-policy exposure. Its discipline is symmetric: it does NOT flag a PassRole combo killed by an explicit Deny, an Action '*' pinned to one bucket, an sts:AssumeRole whose target does not trust back, a mutation kit capped by a permissions-boundary Deny, or a cross-account assume sealed by an unsatisfiable Condition. Reports findings with severity and a fix, then names what a single principal's policies cannot answer (the privileges of a passed/assumed role, the permissions boundary, the org SCPs). Use when asked to audit an IAM policy, role, or user for escalation, over-broad grants, or "can this principal become admin." Vendor-neutral; runs offline against the policy JSON with no Anyshift account.

mkdir -p ~/.claude/skills/iam-deceptive-escalation-auditor
curl -L https://claudskills.com/skills/iam-deceptive-escalation-auditor/SKILL.md \
  -o ~/.claude/skills/iam-deceptive-escalation-auditor/SKILL.md

@misc{iam-deceptive-escalation-auditor-2026,
  author    = {anyshift-io},
  title     = {IAM Deceptive Escalation Auditor [Claude Code skill]},
  year      = {2026},
  publisher = {ClaudSkills},
  url       = {https://claudskills.com/skills/iam-deceptive-escalation-auditor/}
}