tokenpincher@0.5.0

Cut your Claude Code & Codex token bill. One line. No app.

AI Security Review

scanned 5h ago · by lpm-firewall-ai

LPM treats this as warn-only first-party agent extension lifecycle risk. No install-time attack is present. An explicit user command installs a persistent first-party Claude Code hook that receives selected tool outputs; its npx-cache fallback resolves an unpinned latest package version at later hook execution.

Static reason

No blocking static signals were detected.; previous stored version diff introduced dangerous source

Trigger

User runs `tokenpincher init`; Claude Code later invokes its PostToolUse hook.

Impact

Persistent access to matched Claude Code tool output and future execution of the package hook; no confirmed exfiltration or destructive action.

Mechanism

Explicit Claude Code hook installation with npx-latest fallback.

Rationale

This is a real persistent AI-agent extension lifecycle risk, but it is explicit user-invoked first-party setup rather than unconsented install-time hijacking. Warn due to the broad tool-output hook and unpinned `latest` fallback; no concrete malicious behavior was found.

Evidence

package.jsondist/cli.jsdist/hook.jsdist/invocation.jsdist/billing.jsdist/update-check.js~/.claude/settings.json~/.claude/settings.json.tokenpincher-bak~/.tokenpincher/config.json~/.tokenpincher/stats.json~/.tokenpincher/cache~/.tokenpincher/license.json

Network endpoints3

tokenpincher@0.5.0

AI Security Review

Decision evidence

Source & flagged code

Source & flagged code

Findings

Behavioral surface

Source & flagged code