Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add instructions/pieceoffall/jdbg/agents-mdgit clone --depth 1 https://github.com/PieceOfFall/jdbgWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/instructions/pieceoffall/jdbg/agents-md)<a href="https://agentmods.dev/instructions/pieceoffall/jdbg/agents-md"><img src="https://agentmods.dev/badge/instructions/pieceoffall/jdbg/agents-md.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.03638 | $0.03638 |
| Opus 5 | $0.01819 | $0.01819 |
| Sonnet 5 | $0.00728 | $0.00728 |
| Haiku 4.5 | $0.00364 | $0.00364 |
Grade B, and why
jdbg AGENTS.md scanned grade B with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Reads agent configuration directoriesmediumAgent snooping
.claude/, .codex/, .gemini/ hold keys, settings and other credentials a mod has no legitimate need for.
`~/.claude/settings.json`, and the embedded MCP skill to `~/.claude/skills/jdbg/SKILL.md`. Copies of this mod
1 near-identical copy found in the catalogue:
- jdbg CLAUDE.md — 94% identical, 30 lines differ
How it starts
The opening of the file, as written. The whole thing — 191 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md — java-agent-debugger (jdbg)
Coding charter for AI agents working in this repo. Rules and constraints only. For architecture, module layout, command/tool surface, output schema, dependencies, and status, read
DESIGN.md(the factual design-and-implementation reference). This file does NOT duplicate it.
What this project is (one paragraph)
jdbg is a cross-platform Rust CLI (binary jdbg, crate java-agent-debugger, edition 2024) that lets an
AI agent debug Java interactively through the JDK's jdb compatibility backend plus an optional Java JDI
sidecar backend — prompt-aware (never sleep-based), stateful (a background daemon keeps sessions alive across
calls), Windows-first. It is consumed two ways: the CLI and an MCP server (jdbg __mcp, native tool calls
for Claude Code, Codex, and OpenCode). Full detail in DESIGN.md.
Binding constraints (do not violate without asking)
These are settled decisions. Changing them needs explicit user sign-off.
- Keep blocking boundaries where they are settled. The existing
jdbengine, daemon registry, daemon IPC, and per-session command locking remainstd::thread+ channels + blocking IO unless explicitly redesigned.tokiois allowed forrmcpMCP serving and the JDI sidecar transport/lifecycle subsystem. - No temp files, no shell, no sleeps. Commands are written straight to
jdb's stdin. Readiness is detected by reading until the prompt, never by sleeping. There is no shell involved anywhere (no injection surface). - Minimal dependencies. Do not add a crate when
stdsuffices.rmcpis allowed only for MCP protocol/tool serving, not daemon IPC or debugger backend RPC. Deliberately excluded:once_cell,nix,daemonize,portable-pty/conpty, any broad TCP/RPC framework, any windows/winapi crate (usestdraw FFI for the few Win32 calls). Usestd::sync::LazyLockfor one-time regex compilation andstd::os::*for platform bits. - Keep Java sidecar dependencies isolated. Java-only expression/evaluation dependencies belong in
sidecar/jdiand are packaged into the Gradle fat jarjdbg-jdi-sidecar.jar. Do not add a Rust RPC framework or Rust JDWP stack for sidecar work unless the user explicitly approves that redesign. - One daemon per user; one in-flight command per session. The per-session command
Mutexis held across write+wait —jdbis line-oriented and cannot interleave commands. Different sessions run in parallel. - The daemon is the single writer of the on-disk registry (atomic temp-in-same-dir + rename). The CLI only reads it (offline fallback). The in-memory session map is the source of truth while the daemon is alive.
- Keep modules small and single-purpose. If a file grows large it is doing too much — split it.
- Use English for code comments and commit messages. Source/test comments, Rustdoc, git commit messages, and PR/review comments written for the project history should be in English.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 3d ago First seen · 191 lines · 3,638 tokens per session scan B c938a990e193
jdbg AGENTS.md is an instructions file published in the GitHub repository PieceOfFall/jdbg (11 stars, last pushed 1mo ago), licensed Apache-2.0. It adds 3,638 tokens to every session, about $0.0182 per session on Opus 5. A static security scan graded it B with 1 finding (reads agent configuration directories). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other instructions, from other repositories
mcp-jdwp-java CLAUDE.md
Instructions for FgForrest/mcp-jdwp-java, covering claude.md, build commands, run from the reactor root. builds both modules, if it is on path and architecture.
DebugMCP AGENTS.md
Instructions for microsoft/DebugMCP, covering agent guidelines for debugmcp, project overview, architecture, key components and documentation maintenance.
jlink-mcp CLAUDE.md
Instructions for Klievan/jlink-mcp, covering j-link mcp server, building, testing, vscode extension (native mcp integration) and standalone mcp server (for claude desktop / claude code).
debug-that CLAUDE.md
Instructions for theodo-group/debug-that, covering debug-that — debugger cli for ai agents, project overview, tech stack, project structure and commands.
windbg-mcp CLAUDE.md
Instructions for memoryforensics1/windbg-mcp, covering what this is, build & run, build, run (stdio mode — used by mcp clients) and run tests.
x402trace CLAUDE.md
Instructions for fardinvahdat/x402trace, covering claude.md — operating manual for x402trace, project mission, status, hard rules — non-negotiable and strict audit gate (autonomous mode).