ToolTrust
A4/100
samvallad33

vestige

mcpwhitepaper-tr-2026-01

@samvallad33

Think before doing, prove before saving. Vestige is a local, deterministic safety kernel for AI agents. Every memory write and agent action is recorded in Strata, an append-only log, so you can see exactly what your agent did and why.

By samvallad33 | 38 findings | Scanned 10/26/2026 | tooltrust-scanner/v0.3.20 | 656

Repository health

Popularity and maintenance signals alongside the security grade.

View repository
Stars

656

current snapshot

Forks

73

current snapshot

Contributors

17

public GitHub contributors

Last commit

Oct 24, 2026

GitHub repository activity

Last release

Oct 23, 2026

Vestige v5.0.1 — Retroactive Salience Backfill returns

Release version

v5.0.1

Latest GitHub release

Stars trend

Daily star additions · History data pending

—

No observed star events yet.

——

Pull request activity

Weekly PR events · History data pending

No observed PR events yet.

——
RustAGPL-3.0Other11/mo npm downloadsLast scan Oct 26, 2026
12 Low26 Info

Risk Summary

Safe With Normal Controls

Dep Visibility is the main signal, but overall risk remains within an acceptable range.

Potential impact: This finding indicates the tool should be reviewed before it is trusted.

Recommended action: No high-risk findings were detected in this scan, but you should still apply least-privilege defaults and rescan after changes.

Suggested policy: keep this tool behind manual approval, do not allow unattended runs, and re-scan after narrowing risky permissions.

Security Findings (38)

  • LowAS-002

    ⚠️Excessive Permissions ×8

    input schema exposes 19 properties (threshold: 10)

    codebase

    input schema exposes 23 properties (threshold: 10)

    dedup

    input schema exposes 22 properties (threshold: 10)

    ghostlink

    input schema exposes 15 properties (threshold: 10)

    intention

    input schema exposes 24 properties (threshold: 10)

    maintain

    input schema exposes 12 properties (threshold: 10)

    memory_status

    input schema exposes 16 properties (threshold: 10)

    recallsmart_ingest

    Fix: Tool requests broad permissions (exec/fs/network). Validate input parameters using Enums where possible, and restrict file system operations to explicit allowed directories.

  • LowAS-011

    ℹ️Missing Rate-Limit / Timeout ×4

    tool performs network or execution operations but declares no rate-limit, timeout, or retry configuration

    codebasegatherintentionmaintain

    Fix: Declare explicit rate-limit, timeout, and retry configuration for all network and execution tools. Implement exponential back-off and surface resource state to the calling agent.

  • InfoAS-014

    ℹ️Dependency Inventory Unavailable ×18

    Tool did not expose metadata.dependencies or repo_url, so supply-chain coverage is limited.

    causal_walkcodebasededupforgotten_lessongatherghostlinkintentionmaintainmemorymemory_statusprecedentprojectrecallreceiptselftestsession_startsmart_ingestsuppress

    Fix: Review and remediate the identified issue.

  • InfoAS-002

    ⚠️Excessive Permissions ×6

    declared capabilities: code/command execution, filesystem access

    codebase

    declared capabilities: network access, filesystem access

    gather

    declared capabilities: code/command execution

    intention

    declared capabilities: filesystem access, HTTP requests

    maintain

    declared capabilities: filesystem access

    projectsession_start

    Fix: Tool requests broad permissions (exec/fs/network). Validate input parameters using Enums where possible, and restrict file system operations to explicit allowed directories.

  • InfoAS-010

    🔑Insecure Secret Handling ×2

    input parameter "preview_token" accepts a credential (informational; not evidence of insecure handling)

    dedup

    input parameter "allowSecrets" accepts a credential (informational; not evidence of insecure handling)

    smart_ingest

    Fix: Avoid accepting raw credentials as input parameters. Use secret managers (e.g. 1Password CLI, AWS Secrets Manager) and ensure credentials are never logged or stored in agent traces.

Scan this tool yourself

Reproduce this audit locally, integrate into CI, or let your agent audit its own tools.

Install once, then scan any MCP server:

$ curl -sfL https://raw.githubusercontent.com/AgentSafe-AI/tooltrust-scanner/main/install.sh | bash
$ tooltrust-scanner scan --server "npx -y vestige"

Adjust the package name if your npm registry name differs from the tool ID. View source

Add badge to your README

Copy this Markdown to show your ToolTrust grade on GitHub.

[![ToolTrust Grade A](https://raw.githubusercontent.com/AgentSafe-AI/tooltrust-directory/main/docs/badges/grade-a.svg)](https://github.com/AgentSafe-AI/tooltrust-directory)