Scan JSONL exports for memory poisoning, prompt injection and leaked secrets

Mimvo is an open-source Python package for Python 3.11+, published on PyPI. The core package reads JSONL with no extra dependencies.

pip install mimvo

One command reads every record, runs the secrets, injection and poisoning checks, and writes a self-contained HTML report you can forward.

mimvo scan jsonl ./exports/agent_memory.jsonl \  --report report.html

• Each line is a JSON object with content, and optionally id, metadata, embedding and created_at.
• Include embedding so the vector detectors run, and created_at so contradictions between older and newer memories can be found.
• Pass - instead of a path to read from stdin, for example from your own export script.

The same scan from your own code, for a scheduled job or a notebook.

from pathlib import Pathfrom mimvo import Mimvofrom mimvo.scan import JsonlScanSource, render_htmlsource = JsonlScanSource("./exports/agent_memory.jsonl")report = Mimvo().scan(source)report.source = "jsonl:./exports/agent_memory.jsonl"Path("report.html").write_text(render_html(report), encoding="utf-8")

Thirteen finding codes across three checks. Each finding comes with a severity, a recommended action, the evidence (secrets masked), fix steps, and an OWASP and CWE mapping. Details are in the finding code reference.

Fail the job on high-severity findings, send results to GitHub code scanning as SARIF, and put a summary on the run page. See Use in CI for a full GitHub Actions workflow.

mimvo scan jsonl ./exports/agent_memory.jsonl \  --fail-on high \  --sarif mimvo.sarif \  --markdown "$GITHUB_STEP_SUMMARY"

Reads the export line by line. The file is opened read-only.

The default detectors run on your machine with no API key and no model download, so memory text does not leave it. Secret values are masked in every report. Hosted detectors are opt-in.

• ChScan Chroma
• QdScan Qdrant
• pgScan Postgres and pgvector
• PcScan Pinecone
• LCScan LangChain and LangGraph
• m0Scan mem0