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/mitchallen/mcp-hello-server/claude-mdgit clone --depth 1 https://github.com/mitchallen/mcp-hello-serverWrote 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/mitchallen/mcp-hello-server/claude-md)<a href="https://agentmods.dev/instructions/mitchallen/mcp-hello-server/claude-md"><img src="https://agentmods.dev/badge/instructions/mitchallen/mcp-hello-server/claude-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 5.1 | $0.02035 | $0.02035 |
| Opus 5 | $0.01018 | $0.01018 |
| Sonnet 5 | $0.00407 | $0.00407 |
| Haiku 4.5 | $0.00203 | $0.00203 |
Grade A, and why
mcp-hello-server CLAUDE.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 5d 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 — 120 lines — stays where its author put it; the contents beside it link to each section on GitHub.
mcp-hello-server — notes for Claude
A minimal MCP server built with Python / FastMCP — a good starting point for
a new server or a demo. It exposes two tools: server_info (a health/status
check) and greet (a friendly greeting in one of a handful of languages,
defaulting to English). Built with uv, FastMCP, pytest, and
make; multi-stage Docker on a distroless Chainguard/Wolfi Python base
(cgr.dev/chainguard/python) for a near-zero-CVE image. It was scaffolded from the
sibling random-mcp-server by stripping every tool down
to server_info and adding the greet demo tool.
Layout
src/mcp_hello_server/greetings.py— greeting data (GREETINGS), language resolution (names, aliases, ISO codes), and thegreet()builder.src/mcp_hello_server/server.py— the FastMCP server (mcp), its tools (server_info,greet), and themain()console-script entry point (mcp-hello-server).tests/— pytest suite.test_server.pydrives the tools through an in-memory FastMCPClient(no network, no subprocess);test_greetings.pyunit-tests the resolver/builder;test_bdd.py+tests/features/*.featureare a pytest-bdd layer.
Conventions
- Dependencies / venv: managed by uv.
make installrunsuv sync(creates.venv).uv.lockis committed and the Dockerfile installs--frozenfrom it, so runmake lock(oruv lock) wheneverpyproject.tomldeps change. - Running:
make run(stdio, the default MCP transport),make run-http(streamable HTTP onPORT, default 8000),make dev(FastMCP Inspector). The transport is chosen byMCP_TRANSPORT(stdio|http|sse). - Tests:
make test→uv run pytest. - Adding a tool: give it a
titleand MCPannotations. Both current tools are pure reads (readOnlyHint: True,openWorldHint: False); a tool that mutates state or reaches the network should flip those and adddestructiveHint/idempotentHint, which are only meaningful whenreadOnlyHintis false. - Adding a language: add a row to
GREETINGSingreetings.py(and, optionally, an alias / ISO code in_ALIASES).server_inforeports the supported set automatically. - Docker: the image defaults to HTTP transport (
MCP_TRANSPORT=http,HOST=0.0.0.0,PORT=8000) so it's reachable on a published port. The base is distroless Chainguard/Wolfi Python (cgr.dev/chainguard/python), which already runs as the non-rootnonrootuser (uid 65532) and has no shell / package manager. The venv is built on the matching-devimage so its interpreter symlink resolves at runtime.make scanshould report 0 CRITICAL/HIGH. - Trivy version is pinned in two places that must move together. The
Makefile's
TRIVY_VERSION(used asaquasec/trivy:$(TRIVY_VERSION)bymake scan) and theversion:input on everyaquasecurity/trivy-actionstep (image-scan.yml×2,publish.yml,publish-dockerhub.yml,scan-scheduled.yml). Bump all of them in the same change. Dependabot covers neither: itsdockerecosystem reads theDockerfileonly, so it never sees the Makefile tag, and itsgithub-actionsecosystem bumps the action while the scanner binary inside it stays put — the action's own default lags badly (v0.36.0 still defaults to trivy v0.70.0, which is what CI silently ran until the pin landed). Leaving the Makefile on an unpinnedaquasec/trivyis worse than a stale pin: local drifts to whatever:latestresolves to while CI sits on the action default, so a local pass stops predicting the requiredscancheck. - Releasing:
make release(BUMP=patch|minor|major, default patch) bumps the version and opens arelease/vX.Y.ZPR carrying the bump plus theCHANGELOG.mdsection, waits for main's required checks, merges it, then tagsvX.Y.Z— the tag triggers the GHCR, Docker Hub, and PyPI publish workflows (publish,publish-dockerhub,publish-pypi). It then creates the matching GitHub Release (gh release create) using that version'sCHANGELOG.mdsection as the notes (extracted withawk), so the tag, images, package, and Release stay in sync — the Releases page previously drifted because the tag was pushed without a Release object.- Write the new version's
CHANGELOG.mdentry (Keep a Changelog format, top of the file) before running it — the release notes are only as good as that section, and the run aborts if the section is missing. You may leave that edit uncommitted; it rides along in the release PR. Everything else must be committed first. - It must not go back to pushing the bump straight to main. That only ever
worked because
enforce_adminswas off, so the release commit reached main without runningunit/bddand the tag was built from an unchecked commit. Branch protection now enforces admins, so a direct push is refused outright. - The target is re-runnable: if a run stops (checks failed, merge
declined), re-running resumes the existing
release/vX.Y.Zbranch and skips straight to whatever step is left, rather than bumping the version a second time. It also won't re-tag or re-create an existing Release.
- Write the new version's
- Branch protection on
main: required checks areunit,bdd, andscan(the Trivy image gate), withenforce_adminson — nothing reaches main without them, including releases. Becausescanis required, a newly disclosed fixable CRITICAL/HIGH blocks merges until the bump that clears it lands; that's the intended behavior, and Dependabot'suvecosystem opens that bump. - PyPI trusted publishing:
publish-pypi.ymluploads to PyPI via OIDC trusted publishing (no stored token), from thepypiGitHub environment. The PyPI Trusted Publisher must stay constrained to thepypienvironment (ownermitchallen, repomcp-hello-server, workflowpublish-pypi.yml). If it's ever reconfigured, don't leave the environment field blank — an unconstrained publisher lets any workflow/environment in the repo publish, and PyPI will nag you to constrain it. Version badges (PyPI, Python) are cached proxy images (shields.io + PyPI's own image cache) and can lag a release by hours — that's cosmetic, not drift; see the badge note inREADME.md.
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.
- 5d ago First seen · 120 lines · 2,035 tokens per session scan A 4510be233471
mcp-hello-server CLAUDE.md is an instructions file published in the GitHub repository mitchallen/mcp-hello-server (0 stars, last pushed 5d ago), licensed MIT. It adds 2,035 tokens to every session, about $0.0102 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 2026-08-31.
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).
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.
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).
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.
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.
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.