unboundcompute/security-agent-skills/auditing-break-glass-account-trust

Claude Code and agent skill by unboundcompute/security-agent-skills

Trust tier 3
  • community
Source check pending

Checking that the SKILL.md file is still in the repository.

Audit emergency break-glass and privileged-access accounts for the ways their standing power outlives the emergency they exist for: a break-glass account with a static, shared, or never-rotated credential that is a permanent superuser login rather than a sealed last resort, an emergency account excluded from the MFA, conditional-access, or logging controls covering other admins so its use is easy and invisible, a just-in-time elevation that grants more than the task needs or is not revoked when the task ends, an emergency-access path with no approval, time bound, or alerting so it is used routinely, and a break-glass use that triggers no review afterward. Use when a standing emergency identity or an elevation path holds power beyond ordinary admins and the controls around when and how it is used are the boundary. The break-glass credential or elevation request is the source, the unbounded or unaudited privileged access is the sink, and the missing rotation, control coverage, or use-time alerting is the bug.

Install

npx vskill@latest install unboundcompute/security-agent-skills/auditing-break-glass-account-trust

SKILL.md

Auditing break-glass account trust: the last-resort key must be sealed, watched, and rare

A break-glass account exists for the emergency where normal admin access is broken, the identity provider is down, everyone is locked out, so it holds the highest privilege and, by design, bypasses some of the controls that gate everyone else. That combination, maximum power plus fewer gates, is precisely why it is dangerous when its own controls are loose. The account is supposed to be a sealed last resort: strong unique credential, rotated after each use, covered by logging and alerting, used rarely and reviewed every time. The failures are where it stops being sealed. A static, shared, or never-rotated credential turns the emergency key into a permanent superuser login anyone who once saw it can use. An emergency account excluded from the multi-factor, conditional-access, and logging controls that cover other admins is both easy to use and invisible when used. A just-in-time or privileged-access-management elevation that grants more than the task needs, or is not revoked when the task ends, leaves standing privilege behind a temporary label. An emergency path with no approval, no time bound, and no alerting gets used routinely instead of exceptionally. And a break-glass use that triggers no review means the one time it mattered, nobody looked. The audit checks that the emergency identity is sealed, watched, minimally scoped, and reviewed, so its power is available in a real emergency and inert otherwise. You audit this by asking who could use it right now, whether anyone would notice, and whether it would ever be un-elevated.

When to use

  • A system has a break-glass or emergency-access account, or a just-in-time privileged-elevation path, that

holds power beyond ordinary admins.

  • The emergency credential may be static, shared, or never rotated, or the account may be excluded from MFA,

conditional-access, or logging controls.

  • An elevation may over-grant, may not be revoked when the task ends, or a break-glass use may trigger no

alerting or review.

Scope check

Test break-glass and privileged-elevation controls only against accounts and platforms you own or are authorized to assess, with test accounts. Exercising an emergency account or elevation uses real high privilege, so use test identities and never log in as, elevate through, or alter a break-glass account that is not yours. If you can't name the authorization, stop.

The loop

  1. Establish how the emergency identity is meant to be sealed and watched first. Name the break-glass and

elevation paths, what credential each uses and how it rotates, which controls (MFA, conditional access, logging, alerting) cover it, what scope an elevation grants, and how each use is approved and reviewed. This is the false-positive killer: an emergency account with a strong unique credential rotated after use, covered by logging and alerting, elevations scoped to the task and revoked when it ends, and every use reviewed is behaving correctly. Name the intended seal, then test each control.

  1. Test the credential seal. Examine the break-glass credential: whether it is strong and unique or static,

shared, or reused, and whether it is rotated after each use and stored under strict control (split knowledge, sealed vault). A credential anyone who once saw it can still use, or one never rotated, is a permanent superuser login rather than a sealed last resort.

  1. Test control coverage. Confirm the emergency account is subject to the same logging and alerting as every

other admin, and understand exactly which gates it bypasses and why. An emergency account excluded from logging is invisible when used; one excluded from MFA and conditional access with no compensating alerting is both easy to use and unnoticed, which is the opposite of a watched last resort.

  1. Test elevation scope and revocation. For a just-in-time or privileged-access elevation, confirm it grants

only what the task needs and is revoked when the task ends or the time window closes. Elevate, complete a task, and check the privilege is actually removed. An elevation that over-grants, or that leaves privilege standing after the task, is permanent power behind a temporary label.

  1. Test approval, time bound, and alerting on use. Confirm using the emergency path requires approval where

feasible, is time-bounded, and fires an immediate alert and a recorded justification. Use the path and check that an alert reaches someone and a review is prompted. An emergency path with no approval, no expiry, and no alert gets used routinely and silently instead of exceptionally and visibly.

  1. Confirm and record. Confirm with test accounts by using a static or unrotated break-glass credential,

acting through an emergency account that logs or alerts nothing, keeping privilege after an elevation should have ended, or exercising an emergency path that prompts no review, without touching real emergency accounts. Kill the lead if the credential is strong, unique, and rotated after use, the account is logged and alerted, elevations are scoped and revoked, and every use is reviewed. Record the break-glass credential or elevation request, the unbounded or unaudited privileged access, and the missing rotation, control coverage, or use-time alerting.

Where break-glass trust leaks

  • Static or shared credential. A break-glass password that is shared, reused, or never rotated is a permanent

superuser login for anyone who ever saw it, not a sealed last resort.

  • Excluded from controls. An emergency account left out of logging is invisible when used, and one left out

of MFA and conditional access with no compensating alerting is easy and silent to abuse.

  • Over-granting or non-revoked elevation. A just-in-time elevation that grants more than the task needs, or

is not revoked when the task ends, leaves standing privilege behind a temporary label.

  • No approval, time bound, or alert. An emergency path with no approval, no expiry, and no alerting gets used

routinely and unnoticed instead of exceptionally and visibly.

  • No post-use review. A break-glass use that triggers no review means the highest-privilege action in the

system happens with nobody looking afterward.

Worked example (a confirm and a kill)

Confirm. A cloud tenant keeps a break-glass global-admin account with a static password stored in a shared
document, excluded from MFA and conditional access "so it always works in an emergency," and its sign-ins raise
no alert. On a test tenant, the account logs in with the shared static credential, performs privileged actions,
and nothing alerts or is reviewed, so it functions as an always-available, unwatched superuser rather than a
sealed last resort. Confirmed unbounded and unaudited break-glass access via static shared credential and
excluded controls, high, remediation = store the credential under split knowledge in a sealed vault and
rotate it after each use, keep the account covered by logging with an immediate alert on any sign-in, and
require a recorded justification and post-use review for every use.
Kill. The break-glass credential is strong and unique, held under split knowledge in a sealed vault, and
rotated after every use; the account is fully logged and any sign-in fires an immediate alert with a required
justification; just-in-time elevations are scoped to the task and revoked when it ends; and every use triggers
a review. Using the account is possible in a real emergency but never quiet or routine, and elevation never
outlives its task. Killed, kill_reason = "the emergency credential is strong, sealed, and rotated after
use, the account is logged and alerted, elevations are scoped and revoked, and every use is reviewed; the
last-resort power is available but bounded and watched."

Rationalizations to reject

  • "It has to work when everything else is down." → Available is not the same as unwatched or static; a sealed,

rotated credential with alerting still works in an emergency and is not a standing superuser the rest of the time.

  • "We exclude it from MFA so we never get locked out." → An emergency account excluded from controls with no

compensating alerting is the easiest and quietest account to abuse; add compensating monitoring for every gate it bypasses.

  • "Only a few people know the password." → A shared static password known to a few and never rotated is known

to everyone who ever left; make it unique, sealed, and rotated after each use.

  • "The elevation is temporary." → Confirm the privilege is actually revoked when the task or window ends; an

elevation that is not un-done is permanent power with a temporary name.

  • "We would notice if it were misused." → Without logging, alerting, and post-use review you would not; verify

a use actually reaches someone and prompts a review.

Executing this in practice

You need the break-glass and elevation paths, each credential's strength and rotation, which controls cover or exclude the account, the scope and revocation of elevations, and the approval, time bound, alerting, and review around each use. With test accounts, use the credential, act through the account, elevate and complete a task, and check what is logged, alerted, revoked, and reviewed. Reading the emergency-access procedure and platform configuration shows the intended seal; an account that is static, silent, over-granted, or un-reviewed shows where the seal is broken.

Related

  • hunting-iam-privilege-escalation-paths - a loosely controlled break-glass account or elevation is a

high-value escalation target; that skill traces the paths that reach it.

  • auditing-mfa-enrollment-and-reset-abuse - the controls a break-glass account often bypasses; whether the

bypass is compensated is shared ground.

  • mapping-service-account-impersonation-chains - emergency and highly privileged identities are prime nodes in

an impersonation chain; that skill maps where their authority leads.

  • reviewing-secrets-manager-access-policy-trust - where the sealed break-glass credential should live and who

may retrieve it is that skill's subject.

the unbounded or unaudited privileged access, evidence = the missing rotation, control coverage, or use-time alerting.

Security

  • Security scanPassed, 90 of 100Pattern checks on the skill's files.
  • Model reviewPassed

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

Repository stars
5
Repository forks
2
7-day trend score
9

Stars and forks are for the whole GitHub repository, not this skill alone. Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/unboundcompute/security-agent-skills/auditing-break-glass-account-trust/badge)](https://verified-skill.com/skills/unboundcompute/security-agent-skills/auditing-break-glass-account-trust)

Something wrong with this skill or its page? Tell us.