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/minghinmatthewlam/agent-guards/agents-mdgit clone --depth 1 https://github.com/minghinmatthewlam/agent-guardsWhat 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 | $0.00475 | $0.00475 |
| Opus 5 | $0.00237 | $0.00237 |
| Sonnet 5 | $0.00095 | $0.00095 |
| Haiku 4.5 | $0.00047 | $0.00047 |
Grade A, and why
agent-guards 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 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.
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 — 35 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Agent Operating Guidelines
Listen: these rules are persistent constraints, not initial suggestions. Apply them for the full session.
Workflow
- Clarify before acting when the task is ambiguous, high-risk, or has multiple viable approaches. Define success criteria first.
- Verify premises through source code before designing around them: Do not inherit unverified claims — platforms evolve.
- Before substantial or judgment-heavy implementation, give the user a concise plan focused on the approach, important decisions or trade-offs, and verification. Give them a chance to adjust it; skip the pause for routine, low-risk work.
- For non-trivial work, plan verification up front with
self-test. If no self test setup, build it too. - Do not mark work complete before self testing on user level surface.
Output
- Default to concise, status-first replies. The human should be able to scan the result in seconds.
- Put detail in artifacts, diffs, logs, proof paths, or follow-up answers instead of long paragraphs. User will ask for follow up deep dives if wanted.
- For substantial work, lead with status, result, evidence, decision needed, next action, and residual risk.
- Use priority tags (
P0,P1,P2) for findings, blockers, risks, and options, but only include the highest-signal items.
Code
- KISS: Use the simplest architecture that meets the current goal. The user and future agents must be able to understand what happens and why; simple control flow is easier to verify, maintain, and extend.
- Prefer one clear path. Fail fast with a clear error; add a fallback only after a real failure shows it is needed.
- When refactoring, remove old and duplicate paths instead of keeping both.
- Fix root causes, not symptoms.
Git
- Make granular, focused commits during the work, not only at the end.
Philosophy
- Always root your replies about codebases with source code and files, not your intuition or assumption without confirming in source.
- Success criteria first. If “done” is unclear, stop and clarify before executing.
- Keep the human focused on product context, trade-offs, and decisions that require judgment.
- If confidence is below 85%, clarify rather than guessing.
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 · 35 lines · 475 tokens per session scan A b87ccf9086e7
agent-guards AGENTS.md is an instructions file published in the GitHub repository minghinmatthewlam/agent-guards (36 stars, last pushed 14d ago), licensed MIT. It adds 475 tokens to every session, about $0.0024 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
vscode buildNext.instructions.md
Working notes and architecture documentation for the new esbuild-based build system in build/next. Use when making changes to the new build pipeline (transpile/bundle commands, NLS plugin, source-map handling, resource copying, or self-hosting watch tasks).
spec-kit AGENTS.md
AGENTS.md instructions for github/spec-kit, covering agents.md, about spec kit and specify, quickstart — add a new integration in 5 steps, integration architecture and integrationmanifest — file tracking.
codex AGENTS.md
AGENTS.md instructions for openai/codex, covering rust/codex-rs, the codex-core crate, code review rules, crate api surface and model visible context.
langchain AGENTS.md
AGENTS.md instructions for langchain-ai/langchain, covering global development guidelines for the langchain monorepo, corridor security analysis, project architecture and context, monorepo structure and development tools & commands.
vscode oss-third-party-notices.instructions.md
Instructions for microsoft/vscode, covering vs code oss third-party-notices pipeline, architecture, pipeline flow in ci, applying the notice (cutover) and fallback chain (never fail the build).
next.js AGENTS.md
Instructions for vercel/next.js, covering next.js development guide, codebase structure, monorepo overview, core package: packages/next and other important packages.