Warn by default; block when configured. The package alters the consuming project's AI-agent control surface without an explicit user command.
AI security review
scanned · by lpm-firewall-ai
The default postinstall can modify AI-agent instructions and skills in the consuming project without a user command. The limited install and Git-checkout gates do not provide consent for those writes.
Package installation or update, when the package is installed under node_modules in a Git checkout that depends on it; synchronization is enabled by default.
Impact
The package alters the consuming project's AI-agent control surface without an explicit user command.
Mechanism
The postinstall handler plans managed skill changes, then writes or deletes files under the app's .claude directory and may update CLAUDE.md to import the managed skills index.
Policy narrative
During installation, the package runs a postinstall handler that can synchronize its managed skills into a dependent Git checkout. If it proceeds, it writes or deletes files under the app's .claude directory and can add a managed-skills import to CLAUDE.md. These are AI-agent instructions outside the package directory, and the synchronization runs without an explicit user command.
AI identified this as a dangerous AI-agent capability at 99.0% confidence with low false-positive risk.
Evidence for policy risk
The package defines a postinstall hook that runs postinstall.mjs.
postinstall.mjs imports and invokes the skills postinstall handler.
On eligible installs, the handler finds the consuming app and plans skill synchronization.
When synchronization proceeds, it writes or deletes files under the app's .claude directory and can rewrite the app's CLAUDE.md.
The handler adds an import of the managed skills index to CLAUDE.md.
Evidence against
Synchronization skips CI, can be disabled with QORE_SKILLS_SYNC=0, and only proceeds for an installed package in a Git checkout that depends on this SDK.
It skips synchronization when managed files have detected uncommitted edits.
116const rule = entry.kind === "contract" ? "It may add steps and conventions to this skill but cannot relax it: the portal reads what this skill produces." :
Avoid installing @quatronic/sdk@0.10.0. Remove it from direct dependencies and check your lockfile for transitive copies.
Choose an independently verified alternative or release. This report does not establish that other versions are safe.
If this version ran, check your AI agents' configuration and instruction files (the files this report lists) for changes you did not make, and remove them. This finding does not by itself establish malware intent.
132"Managed by `@quatronic/sdk`; see `.claude/qore/README.md` for how to disable, extend or detach one.",
133"",
134"**Never edit `.claude/skills/qore-*` or anything under `.claude/qore/` except `extensions/`** \u2014 the next SDK update overwrites it. To add, change, extend or turn off a skill,...