registry / stratless / 0.3.3

stratless@0.3.3

Give your AI coding assistant a model of who you are — read from your own history — so it stops making you feel stupid. Local, private, no account.

AI Security Review

scanned 3d ago · by lpm-firewall-ai

LPM treats this as warn-only first-party agent extension lifecycle risk. The explicit CLI can archive Claude Code transcripts, generate a user profile, and modify Claude Code configuration. `init --auto` persists a package-owned Stop hook that invokes `stratless update`; no install-time mutation occurs. No concrete malicious exfiltration or remote payload path was found.

Static reason

No blocking static signals were detected.

Trigger

User runs `stratless init`, `stratless init --auto`, or `stratless update`.

Impact

User-consented persistence and access to local Claude transcript content; the borrowed Claude CLI may process generated prompts.

Mechanism

Explicit local transcript processing with optional Claude Code Stop-hook/profile setup.

Rationale

No source evidence establishes malicious behavior, but the package has an explicit opt-in persistent Claude Code hook and profile injection control surface. Downgrade to warn under the firewall policy for first-party AI-agent extension lifecycle risk.

Evidence

package.jsondist/index.jsdist/init.jsdist/sink.jsdist/transcript.jsdist/claude.jsdist/stream.jsdist/notify.js~/.claude/settings.json~/.claude/projects/**/*.jsonl~/.stratless/archive/**~/.stratless/profile.txt~/.claude/HUMAN.md~/.claude/CLAUDE.md

Network endpoints

stratless@0.3.3

AI Security Review

Decision evidence

Source & flagged code

Source & flagged code

Findings

Source & flagged code