Three Public Vulnerabilities. Chained.

  • by

Full opportunity report: Three Public Vulnerabilities. Chained. on ThorstenMeyerAI.com — validation score, market gap, and execution plan.

TL;DR

On May 11, 2026, attackers exploited a chain of three publicly documented vulnerabilities to compromise TanStack npm packages within six minutes. The attack highlights the speed at which public research can be weaponized and the need for faster defense deployment.

On May 11, 2026, attackers exploited a chain of three publicly documented vulnerabilities to publish 84 malicious versions of TanStack npm packages within six minutes, without stealing tokens or compromising the publish workflow itself. This incident underscores how publicly available security research can be rapidly weaponized, outpacing defenders’ mitigation efforts.

The attack involved chaining three vulnerabilities: the pull_request_target “Pwn Request” pattern, GitHub Actions cache poisoning across trust boundaries, and OIDC token extraction from runner memory. Each was previously documented in public security research, but their combination enabled the breach. The attacker created a fork of TanStack/router on May 10, then inserted malicious code through a crafted commit on May 10 at 23:29. The malicious pull request was opened on May 11 at 10:49, triggering workflows that allowed the attacker to mint an OIDC token in memory and exfiltrate credentials via an encrypted messaging network, without stealing tokens or compromising the package registry directly.

The attack was executed using operational tradecraft based on publicly available research findings from GitHub Security Lab, Adnan Khan, and StepSecurity, all published in 2024 and 2025. The chain of vulnerabilities bridges trust boundaries within the CI/CD pipeline, enabling malicious code to reach the npm registry. The incident affected over 160 packages in the ongoing Mini Shai-Hulud campaign, which includes other high-profile compromises like Mistral AI and UiPath.

Three Public Vulnerabilities. Chained.

DISPATCH / MAY 2026
SECURITY · TANSTACK FORENSICS · 3 PUBLIC VULNS · PART 7
▲ Part 7 · Security
TanStack Forensics · May 2026
Software Security · Part 7 · The TanStack Forensic Case Study

Three public vulnerabilities.
Chained.

The TanStack npm compromise of May 11, 2026 — published research recombined into working tradecraft, weaponized faster than defenders deploy mitigations.

84 malicious versions across 42 packages. Six-minute publish window. No npm tokens stolen. OIDC minted in memory and exfiltrated via Session Protocol. Three vulnerabilities chained — each documented in public research 12-24 months before the attack. Same date as the GTIG zero-day disclosure. The composition is the attack surface.

▲ The research-to-tradecraft compression problem
Three pieces of public research. 12 months between the latest and the attack. Zero novel attacker tradecraft. The defender’s deployment of mitigations runs slower than the attacker’s composition of published research. The TanStack incident is the canonical 2026 empirical example.
— software security · the TanStack forensic case study · part 7 · may 2026
84/42
Malicious versions · 42 packages compromised
Two versions per package · 6-minute publish window · @tanstack/react-router 12M weekly downloads
12mo
Latest published research to attack composition
Adnan Khan cache poisoning May 2024 · tj-actions OIDC extraction March 2025
20min
Publish to external detection · Socket flagged in 6 min
Ashish Kurmi · StepSecurity · GitHub issue #7383 · IOC pattern published immediately
160+
Packages in broader Mini Shai-Hulud campaign · May 2026
TanStack · UiPath · Squawk · Mistral AI · DraftLab · Intercom-client · TeamPCP
MAY 11 2026 19:20:39 UTC · FIRST PUBLISH WAVE · 19:26:14 SECOND PUBLISH WAVE · 6 MINUTES BETWEEN
THREE VULNS PULL_REQUEST_TARGET PWN REQUEST · CACHE POISONING ACROSS TRUST BOUNDARY · OIDC MEMORY EXTRACTION
SAME DATE AS GTIG ZERO-DAY DISCLOSURE · TWO AI-AUGMENTED OFFENSIVE EVENTS ON MAY 11 · REMARKABLE CONFLUENCE
MINI SHAI-HULUD 160+ PACKAGES · TANSTACK · UIPATH · SQUAWK · MISTRAL AI · INTERCOM-CLIENT 361K WEEKLY · SELF-PROPAGATING WORM
SLSA L3 FIRST DOCUMENTED VALID-ATTESTATION NPM WORM · NPM AUDIT SIGNATURES PASSES FOR MALICIOUS PACKAGES
DEFENDER ACTIONS ROTATE EVERYTHING · AUDIT PULL_REQUEST_TARGET · PIN SHAS · MOVE OFF OIDC TO SHORT-LIVED TOKENS
MAY 11 2026 19:20 UTC · 84 VERSIONS / 42 PACKAGES · OIDC IN-MEMORY MINT · SESSION PROTOCOL EXFIL
The structural argument · three known vulnerabilities, none sufficient alone

Each bridges the trust boundary the others assumed.

PR fork code crossing into base-repo cache. Base-repo cache crossing into release-workflow runtime. Release-workflow runtime crossing into npm registry write access. The composition only works because each vulnerability bridges the trust boundary the others assumed.

Three public vulnerabilities chained · each necessary, none sufficient
Every component was documented in public research before the attack. The TanStack postmortem explicitly notes the attacker reused verbatim code (with attribution comment preserved) from prior research disclosures.
▲ Vuln 01
pull_request_target · the Pwn Request pattern
BRIDGES: Fork code → base-repo cache
bundle-size.yml ran pull_request_target for fork PRs and checked out the fork’s PR-merge ref to run a build. Bypasses first-time-contributor approval gate. Author attempted trust split but missed that actions/cache@v5‘s post-job save is not gated by permissions:. Cache scope is per-repo, shared across triggers.
PUBLIC RESEARCHGitHub Security Lab · Preventing pwn requests · years before attack
▲ Vuln 02
GitHub Actions cache poisoning across trust boundaries
BRIDGES: Base-repo cache → release runtime
Malicious payload writes to pnpm-store under key release.yml will compute and look up. Linux-pnpm-store-${hashFiles(‘**/pnpm-lock.yaml’)} — exact match. actions/cache@v5 post-step saves poisoned store to that key. Restored entirely as designed when release.yml next runs on push to main.
PUBLIC RESEARCHAdnan Khan · The Monsters in Your Build Cache · May 2024 · 12 months prior
▲ Vuln 03
OIDC token extraction from runner memory
BRIDGES: Release runtime → npm publish
release.yml declares id-token: write for legitimate npm OIDC trusted publishing. Poisoned cache invokes attacker binaries: locate Runner.Worker via /proc/*/cmdline, dump memory via /proc//maps + /proc//mem, extract OIDC token, POST to registry.npmjs.org. Bypasses workflow’s Publish Packages step entirely.
PUBLIC RESEARCHStepSecurity · tj-actions/changed-files compromise · March 2025 · verbatim script reused

The attacker did not invent novel tradecraft. They recombined published research. Verbatim Python script — attribution comment preserved — from the March 2025 tj-actions disclosure. Every defensive research publication becomes attacker reference material within 12-24 months.

Forensic chronology · 28 hours from fork to detection

May 10 17:16 fork. May 11 19:50 detection.

From the attacker creating a renamed fork (deliberately evading fork-list searches) through the cache poisoning phase, the detonation phase, and the rapid external detection by Ashish Kurmi at StepSecurity. The TanStack postmortem published the complete root cause analysis publicly within hours.

Verified forensic timeline · May 10-11, 2026 (UTC)
All times UTC. Cross-referenced from TanStack postmortem (Tanner Linsley), StepSecurity analysis (Ashish Kurmi), Socket research, and the GitHub Security Advisory GHSA-g7cv-rxg3-hmpx.
May 10 17:16Setup
Attacker creates renamed fork to evade fork-list searches
github.com/zblgg/configuration — fork of TanStack/router deliberately renamed. Defender enumerating forks of TanStack/router by name would not surface this. Operational tradecraft — not novel, but deliberate.
SETUP
PHASE
May 10 23:29Malicious commit
Forged “claude” commit lands · fabricated identity
Commit 65bf499d authored by fabricated identity claude (NOT real Anthropic Claude). [skip ci] prefix suppresses CI on push. Adds packages/history/vite_setup.mjs — ~30,000-line bundled JS payload.
POISON
PREP
May 11 10:49PR opens
PR #7378 opens · pull_request_target auto-runs
“WIP: simplify history build” opened by zblgg. bundle-size.yml + labeler.yml auto-run via pull_request_target. No first-time-contributor approval — pull_request_target bypasses that gate. pr.yml blocked.
VULN 01
TRIGGER
May 11 11:11Force-push
Malicious commit lands on PR head · bundle-size.yml executes payload
11:11 force-push lands 65bf499d on PR head. bundle-size.yml’s benchmark-pr job checks out refs/pull/7378/merge, runs pnpm install + pnpm nx run @benchmarks/bundle-size:build. Executes fork-controlled vite_setup.mjs.
PAYLOAD
EXEC
May 11 11:29Cache poisoned
Cache entry saved · poison persists across runs
Cache Linux-pnpm-store-6f9233a50def742c09fde54f56553d6b449a535adf87d4083690539f49ae4da11 (1.1 GB) saved for TanStack/router, scoped to refs/heads/main. Keyed to match what release.yml will compute on next push.
VULN 02
ACTIVE
May 11 11:31Cleanup
Attacker covers tracks · PR appears clean
Force-push reverts PR to current main HEAD (b1c061af). Visible PR diff is 0-file no-op. PR closed and branch deleted in same minute. Cache poison persists. PR appears benign in retrospective review.
EVASION
May 11 19:20:39Detonation
First publish wave · OIDC minted in memory
Manuel merges PR #7369 → release.yml runs (run 25613093674). Poisoned cache restored. Malware locates Runner.Worker via /proc/*/cmdline, dumps memory, extracts OIDC token, POSTs to registry.npmjs.org. Bypasses defined Publish Packages step entirely.
VULN 03
EXEC
May 11 19:26:14Second wave
Second publish wave · same OIDC mechanism
Manuel merges PR #7382 → release.yml runs again (run 25691781302). Same poisoned cache. Second-version-per-package set published — @tanstack/history@1.161.12 etc. Six minutes between the two publish waves. Workflow status: failure (tests broke; publish still happened).
FULL
BLAST
May 11 19:50Detection
Ashish Kurmi opens issue #7383 · complete IOC fingerprint
StepSecurity researcher Ashish Kurmi opens TanStack/router#7383 with full technical writeup. Socket flagged every malicious version within 6 minutes of publication. External detection community had IOC pattern within minutes. Tanner Linsley receives phone call from Socket.dev.
EXTERNAL
DETECTION
May 11 20:00-21:30Response
Incident response · scope confirmed, hardening shipped same day
War room activated. Manuel removes team push permissions. Tanner emails security@npmjs.com. Comprehensive scan confirms 42 packages, 84 versions. Hardening PR merged same day: bundle-size.yml restructured, repository_owner guards added, third-party action refs pinned to SHAs. GHSA published, CVE requested.
IR
COMPLETE
The broader campaign · TanStack as one node

160+ packages. One worm. Same threat actor.

The TanStack compromise is one node in the broader Mini Shai-Hulud campaign by threat group TeamPCP — the same actor behind LiteLLM PyPI (March 2026), Bitwarden CLI npm, SAP CAP npm, and Lightning PyPI (April 30, 2026). Self-propagating worm pattern. First documented npm worm with valid SLSA Build Level 3 attestations.

Mini Shai-Hulud campaign · operational continuity
Same threat actor (TeamPCP / UNC6780) iterating on the same playbook across multiple package ecosystems. Self-propagation via maintainer search + OIDC trusted-publishing abuse.
160+
Packages compromised
May 2026 wave
12M+
@tanstack/react-router
weekly downloads
361K
intercom-client weekly
compromised May 12
29hr
Worm propagation
fork → detection
▲ Current victim organizations · May 2026 wave
TanStack · UiPath · Squawk · Mistral AI · DraftLab · Intercom-client — packages from completely separate maintainer organizations propagated through the worm’s maintainer-search mechanism: registry.npmjs.org/-/v1/search?text=maintainer: → republish with same injection. Active operational campaign as of May 12, 2026.
▲ TeamPCP operational history · prior compromises in same playbook
LiteLLM PyPI · March 24, 2026 · versions 1.82.7 + 1.82.8 · SANDCLOCK credential stealer in 3.4M daily downloads
Bitwarden CLI npm · earlier 2026 · same playbook
SAP CAP npm · earlier 2026 · enterprise blast radius
Lightning PyPI · April 30, 2026 · versions 2.6.2 + 2.6.3
▲ The SLSA Build Level 3 problem · structural defensive breakage
First documented npm worm that produces validly-attested malicious packages. Attacker used stolen OIDC tokens with the legitimate Sigstore stack to produce valid Build L3 attestations. npm audit signatures passes for the malicious TanStack versions. Provenance attestations alone do not validate package safety — defense-in-depth applies.
Indicators of compromise · defender hunt references

IOCs · copy-pasteable for hunting queries.

The TanStack postmortem published comprehensive IOCs. Defenders should hunt for these across their environments. The attacker forged a “claude” identity using claude@users.noreply.github.com — not the real Anthropic Claude Code GitHub App. This identity-confusion tactic deserves specific attention in git-log audits.

Verified IOCs · TanStack-published indicators of compromise
Hunt these across npm dependency manifests, GitHub Actions cache, runtime exfiltration logs, and git commit history. Audit ~/.claude/ and .vscode/ for persistence hooks.
▲ OPTIONAL DEPENDENCY INJECTION · IN PACKAGE.JSON
Malicious optionalDependencies entry · resolves to fork payload
“@tanstack/setup”: “github:tanstack/router#79ac49eedf774dd4b0cfa308722bc463cfe5885c”
Orphan commit lives in attacker fork’s git object store · reachable via TanStack/router URL. prepare lifecycle hook executes bun run tanstack_runner.js && exit 1 on install — payload runs, then optional dep “fails” gracefully.
▲ POISONED CACHE KEY · GITHUB ACTIONS
GitHub Actions cache · 1.1 GB poisoned pnpm store
Linux-pnpm-store-6f9233a50def742c09fde54f56553d6b449a535adf87d4083690539f49ae4da11
Scoped to TanStack/router refs/heads/main. Saved by attacker workflow, restored by legitimate release.yml. File: router_init.js (~2.3 MB, package root, not in files array). Also: tanstack_runner.js per Socket analysis.
▲ EXFILTRATION NETWORK · SESSION PROTOCOL
Session/Oxen messenger exfil · E2E encrypted, no C2 to block
filev2.getsession.org · seed1.getsession.org · seed2.getsession.org · seed3.getsession.org
End-to-end encrypted Session Protocol exfil — no attacker-controlled C2. Blocking by IP/domain is the only network mitigation. 2nd-stage payloads: https://litter.catbox.moe/h8nc9u.js, https://litter.catbox.moe/7rrc6l.mjs. Secondary exfil via legitimate-looking GitHub GraphQL API traffic.
▲ FORGED IDENTITY · NOT REAL ANTHROPIC CLAUDE
Fabricated “claude” commit author · git-log search recommended
claude
Not the real Anthropic Claude Code GitHub App. Fabricated GitHub no-reply identity exploiting display-name confusion. Recommended search: git log –all –author=claude@users.noreply.github.com across all repos. Force-push revert if found.
▲ PERSISTENCE HOOKS · SURVIVES REBOOTS
Persistence in ~/.claude/ and .vscode/tasks.json
router_runtime.js · setup.mjs · settings.json hooks · tasks.json entries
Attacker accounts: zblgg (id 127806521) · voicproducoes (id 269549300 · account created 2026-03-19 — fresh account, public repos named “A Mini Shai-Hulud has Appeared”). Attacker fork: github.com/zblgg/configuration (renamed). Workflow runs: 25613093674 · 25691781302.
Defensive priorities · three audiences

Installed it? Rotate. Maintain packages? Audit.

Three response tracks. If you installed an affected version on May 11: treat your host as compromised. If you maintain OSS with similar workflow patterns: audit pull_request_target immediately. If you consume the npm ecosystem at enterprise scale: deploy install-time monitoring and lockfile pinning.

Three-audience defensive response · prioritized actions
Recovery (rotate everything) · prevention (audit + harden) · monitoring (install-time scanning + lockfile pinning).
▲ IF YOU INSTALLED MAY 11
Rotate everything. Treat host as compromised.

Rotate AWS, GCP, Azure, Kubernetes service-account tokens, Vault tokens, npm ~/.npmrc, GitHub tokens, SSH private keys
Review GitHub Actions runs after 2026-05-11T19:20Z for unexpected npm publish events
Check outbound connections to filev2.getsession.org · seed*.getsession.org
Check downstream propagation — if your packages were published during a CI run that installed compromised version, those may also be compromised
Audit ~/.claude/ + .vscode/tasks.json · remove router_runtime.js, setup.mjs
git log –all –author=claude@users.noreply.github.com · revert if found
Run npm token list · revoke unrecognized tokens

▲ IF YOU MAINTAIN OSS
Audit pull_request_target. Pin SHAs.

Audit pull_request_target workflows immediately · never check out fork-submitted code without explicit approval gates
Pin third-party action refs to commit SHAs · actions/checkout@8e5e7e5ab8… not @v6
Separate cache scopes for trusted vs untrusted contexts · explicit restore-keys and key patterns
Consider moving from OIDC trusted publisher to short-lived classic tokens with manual review
Add internal alerting on npm publishes · fire on any publish that doesn’t originate from expected workflow step
Audit other repos for the same bundle-size.yml-style pattern
Restrict id-token: write to only the publish step that needs it

▲ IF YOU CONSUME NPM AT SCALE
Install-time scanning. Lockfile pinning.

Deploy npm package monitoring at install time · Socket / StepSecurity / Snyk · Socket flagged TanStack in 6 minutes
Lockfile-pinned dependencies don’t auto-pull new versions · only consumers installing during the publish window were affected
Audit lockfiles for github: URL optionalDependencies · unusual for production deps, exact pattern used here
CI/CD secret rotation automation · 30-90 day schedule regardless of incident status
Treat provenance attestations as one layer, not sole verification · Mini Shai-Hulud produces valid Build L3 attestations on malicious packages
Establish IR playbooks for OSS supply-chain compromise scenarios

Three pieces of public security research. Twelve months between the latest and the attack. Zero novel attacker tradecraft. A competent maintainer team with 2FA and OIDC trusted publishing — compromised through a chain that no individual vulnerability in their stack would have enabled. The composition is the attack surface.

— Software security · the TanStack forensic case study · Part 7 · May 2026
Source dossier · the receipts

732 Bytes to Root · Part 1
The 90-Day Window Closed · Part 2
The Defender’s Counter-Cascade · Part 3
The OAuth Permission Apocalypse · Part 4
ShinyHunters · The New APT Model · Part 5
The Roblox Cheat That Broke Vercel · Part 6
TanStack · Tanner Linsley · Postmortem: TanStack npm supply-chain compromise · May 11, 2026
GitHub Security Advisory · GHSA-g7cv-rxg3-hmpx
Tracking issue · TanStack/router#7383 · opened by ashishkurmi May 11 19:50 UTC
StepSecurity · Ashish Kurmi · TeamPCP’s Mini Shai-Hulud Is Back: A Self-Spreading Supply Chain Attack Compromises TanStack npm Packages
Socket · TanStack npm Packages Compromised in Ongoing Mini Shai-Hulud Supply-Chain Attack · 6-minute flagging time
Aikido Security · Mini Shai-Hulud Is Back: npm Worm Hits over 160 Packages, including Mistral and Tanstack
Cyber Kendra · TanStack Packages Hit by Sophisticated Supply Chain Attack
Adnan Khan · The Monsters in Your Build Cache: GitHub Actions Cache Poisoning · May 2024
GitHub Security Lab · Keeping your GitHub Actions and workflows secure: Preventing pwn requests
StepSecurity · Harden-Runner detection: tj-actions/changed-files action is compromised · March 2025 · verbatim OIDC memory extraction technique reused
TeamPCP operational continuity · LiteLLM PyPI March 24 2026 · Bitwarden CLI npm · SAP CAP npm · Lightning PyPI April 30 2026
Mini Shai-Hulud campaign · Socket supply chain attacks tracking · 160+ packages May 2026 wave
Historical precedent · Shai-Hulud npm worm September 2025 · 500+ versions across hundreds of packages
IOC · OAuth optional dep injection · @tanstack/setup · github:tanstack/router#79ac49ee…
IOC · Cache key · Linux-pnpm-store-6f9233a50def742c09fde54f56553d6b449a535adf87d4083690539f49ae4da11
IOC · Exfil · filev2.getsession.org · seed{1,2,3}.getsession.org · Session Protocol E2E encrypted
IOC · Forged commit author · claude · NOT real Anthropic Claude
IOC · Attacker accounts · zblgg (127806521) · voicproducoes (269549300 · created 2026-03-19)
IOC · Renamed fork · github.com/zblgg/configuration · evades fork-list searches

Colophon · Part 7

Set in Source Serif 4, IBM Plex Sans, & IBM Plex Mono. Security-advisory aesthetic. Free to embed with attribution.

thorstenmeyerai.com

Software security · the TanStack forensic case study · Part 7 of 7 · May 2026

84/42 · 12 mo · 20 min · 160+

Implications of Chain Exploitation for Supply Chain Security

This incident demonstrates that the most impactful supply chain attacks in 2026 are no longer reliant on novel vulnerabilities but on the rapid combination of publicly known flaws. It highlights the challenge for defenders to deploy mitigations faster than attackers can weaponize published research. The breach underscores the need for improved security practices, faster patching, and better detection of chained vulnerabilities in open-source ecosystems, especially given the high competence of the attacker and the pre-existing public research.

Broader Trends in Public Research and Supply Chain Attacks

The May 2026 TanStack attack is part of a wider wave of supply chain compromises driven by publicly available security research. Over the past year, researchers have documented vulnerabilities in GitHub Actions workflows, cache trust boundaries, and token extraction methods—each of which has been exploited in real-world attacks. The incident underscores a persistent gap between research publication and defensive deployment, with attackers weaponizing these findings at a pace that outstrips mitigation efforts. This event coincides with the first AI-built zero-day disclosed by Google Threat Intelligence Group, illustrating the convergence of AI-augmented offensive techniques and existing vulnerabilities.

“The TanStack incident exemplifies how publicly documented vulnerabilities can be chained to produce highly effective, rapid supply chain attacks, revealing a fundamental challenge in defense speed.”

— Thorsten Meyer, security researcher

Remaining Questions About Attack Scope and Mitigations

It is still unclear how widespread the malicious versions have become beyond the initial breach, and whether additional undetected exploitation has occurred. The precise detection and mitigation measures that will effectively prevent similar chained attacks remain under development. Details about whether the attacker maintained persistence or targeted specific repositories are also not yet confirmed.

Next Steps for Defense and Monitoring in Open-Source Ecosystems

Security teams are expected to enhance detection of chained vulnerabilities, implement faster patching workflows, and improve monitoring of CI/CD pipelines. The incident underscores the urgency for community-wide adoption of best practices, such as stricter code review of forks, better control of trust boundaries, and real-time vulnerability scanning. Ongoing forensic analysis aims to determine the full extent of the breach and refine mitigation strategies to prevent future incidents.

Key Questions

How did the attacker chain the vulnerabilities in the TanStack attack?

The attacker exploited three publicly documented vulnerabilities: the pull_request_target pattern, cache poisoning across trust boundaries, and OIDC token extraction from runner memory. Combining these allowed malicious code to reach the npm registry without stealing tokens or directly compromising the publish workflow.

What does this attack reveal about the security of open-source supply chains?

It shows that publicly available research can be weaponized rapidly, making supply chain security dependent not only on technical controls but also on the speed of defense deployment and community awareness.

Are existing mitigations sufficient to prevent future chained attacks?

Current mitigations are insufficient against such complex, chained exploits. The incident calls for faster patching, improved detection, and stricter trust boundaries within CI/CD pipelines.

Will this lead to changes in how open-source projects handle security?

Yes, it is likely to accelerate adoption of stricter review processes, automated vulnerability scanning, and better control of external dependencies and forks.

What role does public research play in offensive cybersecurity?

Public research provides valuable insights that can be weaponized by attackers, highlighting a need for balancing transparency with the potential for misuse and for faster deployment of defenses.

Source: ThorstenMeyerAI.com

Leave a Reply

Your email address will not be published.