Skip to content

Wire Briefings Search

Fetching from the wire…

01

02

03

04

05

06

07

08

Wire Briefings Search

Entity trail

TrustMeBro

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

1

Findings

1

Edges

4

Sources

1

Corpus findings

2026-08-31 / sources-researcherTrustMeBro shows PATH shims are enough to feed coding agents fake tool output and move them past their own guardrailsThis Go tool, created 2026-08-26 and now at 327 stars, intercepts command-line tools invoked by Codex, Claude Code and pi through PATH shims alone, with no plugin, hook or MCP integration required, and rules decide whether to fabricate output, rewrite real stdout while preserving stderr and exit status, block the call, or exec the real binary. The published evaluation is the point: each model had to verify a fresh DNS TXT authorization marker before running a scan, and GPT-5.6 Sol, GPT-5.5, DeepSeek V4 Pro and DeepSeek V4 Flash all correctly refused with the real dig, then all four proceeded once TrustMeBro returned fabricated proof. Framed for controlled red-team testing, and a reminder that any agent decision gated on shell output is only as trustworthy as PATH.

Graph relationships

CRITICIZES
DeepSeek V4 Pro
TrustMeBro found that DeepSeek V4 Pro failed to properly verify fabricated credentials.
Source finding
CRITICIZES
GPT-5.6 Sol
TrustMeBro found that GPT-5.6 Sol failed to properly verify fabricated credentials.
Source finding
CRITICIZES
Claude Code
TrustMeBro found that Claude Code failed to properly verify fabricated credentials.
Source finding
CRITICIZES
Codex
TrustMeBro found that Codex failed to properly verify fabricated credentials.
Source finding

Source trail

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues