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/hwfengcs/dm-code-agent/agents-mdgit clone --depth 1 https://github.com/hwfengcs/DM-Code-AgentWhat 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 | $0.01264 | $0.01264 |
| Opus 5 | $0.00632 | $0.00632 |
| Sonnet 5 | $0.00253 | $0.00253 |
| Haiku 4.5 | $0.00126 | $0.00126 |
Grade A, and why
DM-Code-Agent AGENTS.md scanned grade A with 0 findings 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 2d 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.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 98 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
Guidance for coding agents working in this repository.
Project Intent
DM-Code-Agent is a local-first, auditable code maintenance agent. Favor changes that improve real repository maintenance, traceability, reproducibility, and benchmark quality.
Development Rules
- Keep the core agent readable. Prefer small, explicit modules over large abstractions.
- Ask first: does this have to live in the kernel? If a feature only needs to act at a few
fixed points, it belongs in an extension (
docs/extensions.md), not in a new--enable-xxxflag plus anifbranch inReactAgent. - Never delete original data. Compaction, truncation, and summarization may only append a derived record and skip the original when building context — never overwrite or drop it. This is what makes the agent debuggable, auditable, and ablatable.
- Do not introduce network calls into tests unless they are clearly marked as live-model tests.
- Default tests and deterministic evals must run without API keys.
- Treat session/trace files as potentially sensitive. Full LLM I/O capture must remain opt-in;
--tracestays redacted and--trace/--checkpointmust not point at the same file. - Tool replay that mutates files or runs commands must remain explicit and documented.
- Keep benchmark tasks executable with plain pytest hidden tests.
- When changing benchmark behavior, update tests and docs together.
- Do not claim evaluation numbers that were not actually run. The scoreboard is the bundled coding + maintenance benchmark (13 hidden-test tasks). Real SWE-bench / Docker / cross-model scoring is frozen.
- Do not mix a fix into a pure-refactor commit — it destroys the "eval identical field by field" check that makes the refactor verifiable.
Verification
Before considering a change complete, run:
python -m compileall dm_agent main.py tests
python -m pytest
python -m dm_agent.evals.cli --variant full --task direct_finish
python -m dm_agent.benchmarks.cli --suite maintenance --list
python -m ruff check .
python -m black --check .
python -m mypy dm_agent
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.
- 2d ago First seen · 98 lines · 1,264 tokens per session scan A 7ab5b6837b7e
DM-Code-Agent AGENTS.md is an instructions file published in the GitHub repository hwfengcs/DM-Code-Agent (151 stars, last pushed 6d ago), licensed MIT. It adds 1,264 tokens to every session, about $0.0063 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other instructions, from other repositories
agentscope copilot-instructions.md
Copilot instructions for agentscope-ai/agentscope, covering agentscope code review guide, 1. code quality, 2. [must] code security, 3. [must] testing & dependencies and 4. code standards.
braindb CLAUDE.md
Instructions for dimknaf/braindb, covering braindb — claude instructions, ⚠ tool priority — read this first, it overrides habit, at the start of every session, 1. get always-on rules (behavioral guidelines) and project structure.
vetix AGENTS.md
Instructions for HuTa0kj/vetix, covering vetix — project map, repository layout, workflow, extension points and configuration.
browseruse-agent-bench copilot-instructions.md
Generic code review instructions that can be customized for any project using GitHub Copilot.
agentpool AGENTS.md
Instructions for phil65/agentpool, covering project overview, development commands, installation & setup, install with uv (recommended) and install specific extras.
browseruse-agent-bench CLAUDE.md
Instructions for lexmount/browseruse-agent-bench, covering claude.md, team-wide agent rules, commands, install gotcha and high-level architecture.