Borrowing it
Nothing to install: this file belongs to randomparity/hmc-mcp. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/randomparity/hmc-mcp/main/AGENTS.mdgit clone --depth 1 https://github.com/randomparity/hmc-mcpWrote 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/randomparity/hmc-mcp/agents-md)<a href="https://agentmods.dev/instructions/randomparity/hmc-mcp/agents-md"><img src="https://agentmods.dev/badge/instructions/randomparity/hmc-mcp/agents-md/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/instructions/randomparity/hmc-mcp/agents-md"><img src="https://agentmods.dev/badge/instructions/randomparity/hmc-mcp/agents-md.svg" alt="Reviewed on agentmods" width="80" 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 5.1 | $0.06263 | $0.06263 |
| Opus 5.5 | $0.02505 | $0.02505 |
| Sonnet 5.5 | $0.01253 | $0.01253 |
| Haiku 4.5 | $0.00626 | $0.00626 |
Grade A, and why
hmc-mcp 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 21d 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 — 446 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
Non-interactive shell discipline
All shell commands must be non-interactive. A command that opens an editor, a pager, or a prompt will stall the agent with no recovery path.
git
- Always set
GIT_EDITOR=truewhen a git command may open an editor:GIT_EDITOR=true git rebase --continue GIT_EDITOR=true git commit --amend GIT_EDITOR=true git merge --continue - Prefer
--no-editoncommit --amendandmergewhen the message is unchanged. - Never run
git rebase --continuewithoutGIT_EDITOR=true— the continue step opens the editor to confirm the commit message even when there are no remaining conflicts. - Suppress the pager:
git --no-pager <subcommand>or setGIT_PAGER=cat. - Never squash-merge (
gh pr merge --squash/git merge --squash) unless the PR is documentation-only. Squash merges collapse commit history and breakgit bisect. For all code, test, config, and script changes, use eithergh pr merge --merge(merge commit) orgh pr merge --rebase(rebase merge).- Documentation-only means no changed file's content is asserted by a
test or a
staticgate. Content, specifically: repo-wide scans likejust secretsread every tracked file, and counting those would make the exception dead. - Decide it by reading, not by a pattern or a grep. Both have been tried
here and both are unsound, because the guards reach their files three
different ways: by literal path, by basename
(
tests/test_project_metadata.pyopensROOT / "docs" / "hmc-cli-cheatsheet.md"), and by directory —scripts/gen_tool_reference.pywrites_REPO_ROOT / "docs" / "tools"andscripts/check_adr_numbering.pyglobsdocs/adr/*.md, so no search for a path or a filename can see those at all. Work from this floor instead:- Content-asserted:
docs/tools/(diffed byjust tool-docs-check), every page carrying a generation banner (regenerated and diffed byjust doc-freshness),docs/environment-variables.md(checked field by field byjust env-vars),docs/hmc-cli-cheatsheet.mdanddocs/authorization-audit.md(opened bytests/), ADR 0029's inventory block (parsed bytests/unit/test_public_api.py), andCHANGELOG.md,CONTRIBUTING.md,README.md,SECURITY.md. - Name-asserted only:
just adr-numberingchecks an ADR's filename and H1, never its body — so editing an ADR's prose is not caught by that gate. Most ADRs are in this class, which is why a prose-only ADR edit is one of the few real documentation-only changes here. - Content-unasserted today: the
docs/scorecard-*anddocs/workflow/pages. Test modules cite some of those specs in module docstrings, which asserts nothing about the file.just doc-freshnessdoes read every tracked Markdown file's first line, looking for a generation banner — so a page here can still redden it by opening with something banner-shaped, but its body is unchecked.
- Content-asserted:
- When the answer is not obvious, use
--merge. It costs one merge commit. A wrong--squashis not reversible once it is onmain. - Two shell mechanics, for whatever check you do write over
gh pr diff <PR> --name-only:grep -Evexits 1 when it filters every line out, so a naive… | grep -Ev '…'reports failure on exactly the empty-result case and is fatal underset -eor inside an&&chain; and a failedgh pr diff— wrong number, expired auth — prints nothing, which looks identical to a clean pass. Check its exit status before reading its output, and treat an unanswered query as "not documentation-only". - The policy is prose, not a gate: nothing in the repo enforces it —
rg -ni squashover.github/,tests/,scripts/,justfile, and.pre-commit-config.yamlreturns nothing. The maintainer has intentionally retained squash-merge availability and selected this documented human pre-merge check as the sufficient control; your own reading before the merge is the only control. - A single-parent commit on
mainis not evidence of a squash. This repo has also landed PRs with--rebase, which replays each commit ontomainand preserves the per-commit history the policy protects — seven commits from PR #455 landed that way asaec6125^..f528e94. To tell a squash from a rebase, compare the commit's own diff with the PR's: on a squash they are equal, on a rebase the commit carries only its own slice. This paragraph is forensic — how to read history after the fact. It does not prefer one permitted non-squash strategy over the other: both--mergeand--rebasepreserve the per-commit history this policy protects.
- Documentation-only means no changed file's content is asserted by a
test or a
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.
- 21d ago Changed · +21 lines · +276 tokens per session abbaa6b6df04
- 29d ago Changed 3d5ed4cbb7ed
- 1mo ago Changed · +15 lines · +194 tokens per session f9d020c1b098
- 1mo ago Changed · +6 lines · +61 tokens per session 3d49d7e3eec9
- 1mo ago Changed · +18 lines · +297 tokens per session 6c5b8ba43664
- 1mo ago Changed · +9 lines · +126 tokens per session 8c8b52ae4ec8
- 1mo ago Changed · -15 lines · -220 tokens per session edd07556fe58
- 1mo ago Changed · +1 lines · +25 tokens per session 5172c90e07e1
- 2mo ago First seen · 391 lines · 5,504 tokens per session scan A 2148c109faaf
hmc-mcp AGENTS.md is an instructions file published in the GitHub repository randomparity/hmc-mcp (1 stars, last pushed 18d ago), licensed MIT. It adds 6,263 tokens to every session, about $0.0251 per session on Opus 5.5. 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 2026-08-31.
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).
next.js AGENTS.md
AGENTS.md instructions for vercel/next.js, covering next.js development guide, codebase structure, monorepo overview, core package: packages/next and other important packages.
vscode oss-third-party-notices.instructions.md
Instructions for microsoft/vscode, covering vs code oss third-party-notices pipeline, architecture, pipeline flow in ci, cache eligibility and rollout and applying the notice (cutover).
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 copilot-instructions.md
Copilot instructions for microsoft/vscode, covering vs code copilot instructions, project overview, root folders, core architecture (src/ folder) and built-in extensions (extensions/ folder).
deno copilot-instructions.md
Copilot instructions for denoland/deno, covering deno development guide for github copilot, network access, high level overview, key directories and key files to understand first.