yeet-src

Md Sentry — AI skill for Claude Code

AI community

eBPF integrity monitor for an LLM agent's markdown brain: CLAUDE.md, skills, memory.

How to install Md Sentry

This entry records only its repository, not the path inside it, so there is no exact command to give. Open yeet-src/md-sentry and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Md Sentry does

eBPF integrity monitor for an LLM agent's markdown brain: CLAUDE.md, skills, memory. Tags AGENT vs EXTERNAL edits.

Alternatives in AI

README

`md-sentry`

**A tripwire for the files that tell your agent who it is.** Every create, modify, delete and rename of an agent's markdown brain, caught in the kernel and tagged by whether the agent itself made the change.

Linux: BTF-capable kernel, verified on 6.1 through bpf-next Built with yeet: syscall tracepoints and VFS fentry probes loaded by the yeet daemon Hooks: openat/write/close tracepoints plus fentry on vfs_unlink and vfs_rename GPL-2.0 Chat with the yeet community on Discord

md-sentry watching a fake agent tamper with its own config: 133 changes tagged agent and 15 tagged external, with the protected board showing CLAUDE.md, AGENTS.md, skills and memory all touched by the agent, and the feed showing the injected lines it wrote

**`md-sentry` is a kernel-level integrity monitor for the markdown files that steer an LLM agent: its instructions (`CLAUDE.md`, `AGENTS.md`), its memory, and its skills. Every change is tagged `agent` when it came from the agent's own process subtree and `external` when it came from anything else.**

The question is not "did this file change". `inotify` answers that, and so does `git status`. The question is **who changed it**. An age