Supply Chain Audit logo

Supply Chain Audit

Organization
crafter-station
supply-chain-audit

Read-only audit of a developer machine for npm/PyPI supply-chain compromise. Checks for known IOCs from the 2025-2026 wave — Shai-Hulud 2.0 (Nov 2025), Mini Shai-Hulud / TeamPCP (May 2026), Axios DPRK (Mar 2026), Shai-Hulud 3.0 / keyv-cacheable (Aug 2026), and any future campaigns added to the IOC pack. Scans persistence artifacts (LaunchAgent / systemd unit / Windows Run key / gh-token-monitor / .claude and .vscode droppers), payload files (router_init.js, setup_bun.js, bun_environment.js, Math_Symbol.js, math_init.js), compromised package versions in every node_modules under the user's project roots, C2 / typosquat domain strings (git-tanstack.com, api.masscan.cloud, sfrclak.com, getsession.org, npm-cache.com), malicious commit hashes (79ac49ee), payload SHA256s, optionalDependencies pointing at git refs, and files written during the published attack windows. Produces a PASS/FAIL verdict, IOC checklist, at-risk package list, phase summary, and 48h bake-period remediation. Use this skill whenever the user asks "am I affected by the npm attack", "scan my machine", "check if I'm infected", "is package X compromised", "audit my coworker's machine", mentions Shai-Hulud / TanStack hack / TeamPCP / Mini Shai-Hulud / keyv hack / cacheable hack / pull_request_target compromise / npm worm / axios attack / Bun installer malware / TruffleHog secret theft, asks about IOCs or supply chain risk, or wants to verify a host after a security disclosure. Trigger even when the user uses informal phrasing ("estoy chiveado?", "ya me hackearon?", "this safe?"). Never modifies files.

Overview

Publishercrafter-station
Repositoryskills
Skill namesupply-chain-audit
Stars
114
Forks
15
Bundled files
4
LicenseMIT
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • 4 bundled files

    Scripts, templates, and references the model can read while it works. Files are read-only and never executed.

  • Open source

    Published by crafter-station on GitHub. Read the source before you install it.

Installation

Install the Supply Chain Audit AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/crafter-station/skills/tree/main/skills/supply-chain-audit
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/crafter-station/skills.git /tmp/skills
mkdir -p .claude/skills
cp -r /tmp/skills/skills/supply-chain-audit .claude/skills/supply-chain-audit
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Supply Chain Audit in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Supply Chain Audit on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Supply Chain Audit is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

supply-chain-audit

A read-only forensics scanner for npm / PyPI supply-chain compromise. Runs three IOC phases against the local machine and produces a clean PASS/FAIL verdict.

When to invoke

The user asked about supply-chain risk, a recently disclosed npm/PyPI compromise, whether their machine is affected, or wants to share this check with coworkers. Triggers reliably for both technical phrasing ("scan for IOCs", "audit my host") and casual phrasing ("am I cooked?", "is this safe?", "ya me hackearon?").

How it works

The IOC pack lives in iocs.json — a versioned list of campaigns, each with its persistence paths, payload filenames, payload hashes, C2/typosquat strings, optional-dependency markers, repo-artifact hashes, forged-commit authors, compromised package scopes, and attack windows. The scanner script scripts/scan.sh reads that file and runs four phases:

  • Phase A — persistence: artifacts that survive reboot (LaunchAgent / systemd / Windows Run key / ~/.local/bin shims, dropper files in ~/.claude/setup.mjs / ~/.vscode/setup.mjs, named lock files).
  • Phase B — code & cache: package versions present in any node_modules under the configured project roots, payload filenames anywhere on disk, malicious commit hashes / typosquat domains / payload SHA256s in lockfiles and source, optionalDependencies entries that resolve to a GitHub git ref (the TeamPCP smuggling pattern).
  • Phase C — time window: any file written under any node_modules during a campaign's published attack window. A clean Phase C is the strongest single signal a host avoided exposure.
  • Phase D — repo artifacts & git history: files the malware plants inside project repos (.vscode/tasks.json, .claude/settings.json, setup.mjs — names too generic for a presence check, so FAIL only on exact SHA256 match against repo_artifact_hashes), and commits authored by forged identities from forged_commit_authors (the Shai-Hulud 3.0 pattern: claude <claude@users.noreply.github.com>) in any git repo under the roots.

Every check is find / grep / jq / stat / shasum / git log. The scanner builds shared filesystem inventories once (one walk per artifact class, not one per IOC) and never writes to the target machine outside its /tmp/supply-chain-audit-* workdir.

Invocation flow

  1. Read iocs.json to understand the current campaign list. If --list is requested, summarize the campaigns and stop.
  2. Determine project roots. Default: $HOME, but prefer narrower paths the user mentions (e.g. ~/Clerk, ~/Programming). Ask once if ambiguous and >50GB of code is in scope.
  3. Run bash scripts/scan.sh [root1 root2 ...]. The script emits a structured markdown report to stdout. Capture it.
  4. Parse the report. Verdict logic: only FAIL rows count toward POTENTIALLY COMPROMISED. A FAIL means an unambiguous compromise signal — persistence artifact present, payload filename found, payload SHA256 matched, string IOC in lockfile, optionalDependencies git-ref smuggle, repo-artifact hash match, forged-author commit in git history, or Phase C time-window hit. Inventory rows ("you have a copy of @tanstack/query-core in your tree") are emitted as REVIEW and do not flip the verdict — having @tanstack/* installed in February 2026 says nothing about whether you got the May 2026 malicious versions. Phase C answers that. If FAIL_COUNT > 0 → POTENTIALLY COMPROMISED. Otherwise CLEAN, even if REVIEW > 0.
  5. Present the report to the user following the output format below.

Output format

ALWAYS use this exact structure:

# Supply-chain audit — <hostname> — <timestamp>

## Verdict: CLEAN  (or  POTENTIALLY COMPROMISED)

## IOC checklist
| Campaign | Check | Result |
|----------|-------|--------|
| ...      | ...   | PASS/FAIL/REVIEW/skipped |

## Inventory — packages from compromised scopes (REVIEW, not FAIL)
(only if Phase B found compromised scopes installed; list `name@version`, install date, path)

## Phase summary
- Phase A (persistence): N fail(s)
- Phase B (code & cache): N fail(s), M review(s)
- Phase C (time window): N fail(s) — authoritative exposure signal
- Phase D (repo artifacts & git history): N fail(s)
- Total: P PASS / F FAIL / R REVIEW / S skipped

## Next steps
<remediation block — see references/remediation.md>

When summarizing REVIEW rows to the user, cross-check the install timestamp against the campaign's attack window from iocs.json. If install date is outside the window AND Phase C is clean → these packages are safe, mention briefly. If install date overlaps the window → escalate the row to "needs manual version check against the disclosed malicious version list."

If verdict is POTENTIALLY COMPROMISED, the Next steps block MUST lead with:

Do NOT revoke any GitHub PAT yet. The Mini Shai-Hulud / TeamPCP payload installs a dead-man's-switch that watches the stolen token and runs rm -rf ~/ the moment it sees a 401 from api.github.com. First audit token activity at https://github.com/settings/tokens (look for unfamiliar IPs and User-Agents in the last 72h). Then rotate from a separate clean machine: npm publish tokens, cloud creds (AWS/GCP), SSH keys, Vault tokens. Only revoke the suspect PAT after you've moved off the suspect host.

If verdict is CLEAN, the Next steps block recommends the 48h bake period:

npm config set min-release-age 2            # npm 11+
pnpm config set minimumReleaseAge 2880      # minutes
echo -e "[install]\nminimumReleaseAge = 172800" >> ~/.bunfig.toml   # seconds
yarn config set npmMinimalAgeGate 48h

Full remediation prose lives in references/remediation.md. Read that file when the user asks for more detail.

Updating the IOC pack

When a new campaign is disclosed, append a new entry to iocs.json following the schema. The scanner picks it up automatically — no edits to the skill body needed. Schema reference: see the top of iocs.json itself. Bump the version field to today's date.

If the user is reporting a brand-new campaign that isn't in the IOC pack, offer to add it. Ask for: campaign name, disclosure date, affected scopes/packages, payload filenames, payload hashes (SHA256), C2 / typosquat domains, persistence paths, attack window (UTC). Then write the new entry and re-run the scan.

Platform support

  • macOS: full support (LaunchAgent paths native).
  • Linux: full support (systemd-user paths).
  • Windows: best-effort. Run via WSL or Git Bash. The Windows Run-key check uses reg query if available, otherwise marked skipped rather than pass.

Hard constraints

  • Read-only. Allowed tools: find, grep, jq, stat, ls, shasum, awk, sed -n (print-only), git log (local history reads only, never fetch/pull), reg query (Windows). Forbidden: any redirect that writes outside /tmp/supply-chain-audit-* workdir, any package-manager command that resolves dependencies (npm install, pnpm install, bun install), any network call to api.github.com or registries.
  • No credential touching. Never read ~/.npmrc, ~/.ssh/, ~/.aws/, or environment variables containing token-like strings. The point is to detect malware, not exfiltrate the user's own secrets while doing so.
  • No revocation suggestions on FAIL until the dead-man's-switch warning has been delivered. This is non-negotiable — premature revocation on an infected host wipes ~/.

References

  • iocs.json — versioned IOC pack, the single source of truth.
  • references/campaigns.md — narrative descriptions of each campaign, attribution, technique chains.
  • references/remediation.md — full bake-period setup, post-compromise playbook, credential-rotation order.
  • scripts/scan.sh — the scanner itself. Read it if you need to extend coverage.

Bundled files

The model reads these on demand while the skill is loaded. They are exposed as readable files and are never executed.

Frequently asked questions

What does the Supply Chain Audit AI skill do?

Read-only audit of a developer machine for npm/PyPI supply-chain compromise. Checks for known IOCs from the 2025-2026 wave — Shai-Hulud 2.0 (Nov 2025), Mini Shai-Hulud / TeamPCP (May 2026), Axios DPRK (Mar 2026), Shai-Hulud 3.0 / keyv-cacheable (Aug 2026), and any future campaigns added to the IOC pack. Scans persistence artifacts (LaunchAgent / systemd unit / Windows Run key / gh-token-monitor / .claude and .vscode droppers), payload files (router_init.js, setup_bun.js, bun_environment.js, Math_Symbol.js, math_init.js), compromised package versions in every node_modules under the user's pr...

Why use Supply Chain Audit on TypingMind?

Because you install it once and use it with any model. Supply Chain Audit is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Supply Chain Audit in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/crafter-station/skills/tree/main/skills/supply-chain-audit. TypingMind reads its SKILL.md and bundles its files and installs it as a skill you can enable per chat.

Which AI models can use Supply Chain Audit?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Supply Chain Audit?

As many as you like. As long as a model supports skills, you can use Supply Chain Audit with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Supply Chain Audit AI skill free?

Yes. It is published on GitHub by crafter-station under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇