The trust prompt on your ai coding agent is the whole security model, and attackers get code…
the trust prompt on your ai coding agent is the whole security model, and attackers get code running before it ever appears. a booby-trapped git config line executes in several agents including claude code and codex, and one 4,176-run benchmark had agents follow malicious instructions 66.5% of the time.
the tool writing your code trusts the repo more than you do.
Context
Verified source: Manifold Security's GitSpawn disclosure (updated Sep 1, 2026) and The Hacker News (Sep 2) report eight flaws across seven command-line AI coding agents where a repo's own Git configuration names a command the agent runs on the developer's machine. The Cloud Security Alliance note (Sep 4) says that in several cases this happens before a trust prompt appears.
Manifold says Codex and Cursor were also affected, reported and patched; The Hacker News says four of the flaws were still unpatched at its writing.
The core claim, that a booby-trapped git config can run code in several agents including before any trust prompt, matches Manifold, The Hacker News and the CSA note. The 4,176-run benchmark and its 66.5% rate were not seen in the sources read, so unsupported here, not refuted. in the excerpts read. The closest benchmark found, IssueTrojanBench on arXiv (Jul 22), tests malicious issue requests, a different attack from the git config one. 'The tool trusts the repo more than you do' is the author's opinion.
Related work
- IssueTrojanBench: Benchmarking AI Coding Agents Against Malicious Issue Requests, arXiv ↗Related but a different attack path.
Watch next
- Locate the benchmark behind the 4,176 runs and 66.5% and check what it measured.
Sources
- GitSpawn: a single flaw lets untrusted repos run code in AI coding agents, Manifold Securitymanifold.security
- Malicious .git configs can make Claude, Codex, Cursor and others run attacker code, The Hacker Newsthehackernews.com
- GitSpawn: Malicious Git Configs Hijack AI Coding Agents, Cloud Security Alliancelabs.cloudsecurityalliance.org
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 30 September 2026 at 21:19 IST. Sources are the papers and datasets the note draws on.
View the original post ↗Embed this note
More notes
The air is now being asked to keep its own ledger
the air is now being asked to keep its own ledger: ecmwf’s aifs compo becomes the first ai model to forecast atmospheric composition globally every three hours, cleanair simulates 365 days of pm2.5 over china in ten seconds, and a unified framework maps six pollutants at one kilometer across the whole country. the air now files its own composition report.
read the note →The current is now being asked to draw its own map
the current is now being asked to draw its own map: china’s langya 2.0 predicts six ocean phenomena including internal waves and mesoscale eddies, a deep net called wenhai resolves eddies globally with air sea flux formulas built in, and scripps infers surface currents from the way temperature patterns deform in satellite images. the ocean now files its own circulation report.
read the note →The soil is now being asked to report its own carbon
the soil is now being asked to report its own carbon: a nix color sensor paired with generative data augmentation predicts soil organic carbon without a lab, random forest drives 74 percent of soil health mapping studies, and sentinel 2 tracks five year carbon change across france and italy from 922 samples. the dirt now files its own carbon account.
read the note →