bastionmemory

Name: bastionmemory
Author: Bastion

scan

Agent memory-poisoning scanner

Detect memory poisoning (OWASP ASI06): directives disguised as facts in what an agent remembers, and poison that appears over time.

pip install bastionmemory

Memory/context sibling of bastionskill (skill code) and bastionsupply (MCP tools).

What it does

Scans the memory an agent auto-loads as trusted every session (v0.1: CLAUDE.md + Claude Code auto-memory)
Judges shape, not vocabulary: declarative facts pass, imperatives aimed at future behavior get flagged
Baseline diff catches stealth and delayed poison, the new lines since you last vetted it
Verdicts allow / review / block; --fail-on CI gate + harden feeds the poison to agentbastion as detector templates

bastionmemory scan ./memory --record bastionmemory scan ./memory --since-baseline

Version 0.2.0 · MIT