supplychain

name	supplychain
description	Supply-chain attack category — dependency confusion, typosquatting, package-registry abuse, build-pipeline poisoning, SBOM manipulation.
allowed-tools	Bash Read Write
metadata	{"subdomain":"supply-chain","when_to_use":"supply chain, dependency confusion, typosquat, package, npm, pypi, rubygems, cargo, maven, gradle, sbom, ci, github actions, build pipeline, registry, vendored","tags":"supply-chain, dependency-confusion, typosquatting, ci-cd","mitre_attack":"T1195, T1195.001, T1195.002"}

Supply-Chain Exploitation — Category Overview

This is a routing skill for supply-chain attack research. Sub-skills cover specific techniques against the dependency, build, and distribution layers.

Sub-Skills

Sub-Skill	Covers	When to Load
dep-confusion	Public-registry impersonation of private package names, scoped-package abuse (@org/pkg), internal-name OSINT (lockfiles, GitHub Actions, Dockerfiles, error pages)	Target uses private npm/PyPI/RubyGems/Cargo registries and you've discovered internal package names in public artifacts

Surface Discovery

# Lockfiles often expose private package names that aren't on public registry
grep -hE '"name":|"@.+/' package-lock.json yarn.lock pnpm-lock.yaml | sort -u
# pip
grep -E '^[a-z0-9_-]+' requirements.txt
# go.mod
grep '^\s*[a-z0-9./_-]\+\s' go.mod
# Maven
grep -A1 '<artifactId>' pom.xml | grep -v '^--$'

Decision Notes

• Always confirm scope authorization in writing before publishing any package, even a benign one, to a public registry. Some bounty programs explicitly disallow public-registry submissions.
• Use a beacon-only payload (DNS callback, HTTPS GET) for confirmation; never ship code that mutates the target environment without an explicit follow-up authorization.
• Real damage from confusion attacks is usually post-install scripts (postinstall, setup.py install) — focus PoCs on benign callbacks, not RCE.
• For build-pipeline poisoning (workflow injection, branch protection bypass) see /skills/standard/exploit/ad-exploitation/SKILL.md (CI-as-AD-equivalent paths) and the broader threat-modeling under /skills/shared/opsec/SKILL.md.