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/rmednitzer/relay-shell/claude-mdgit clone --depth 1 https://github.com/rmednitzer/relay-shellWrote 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/rmednitzer/relay-shell/claude-md)<a href="https://agentmods.dev/instructions/rmednitzer/relay-shell/claude-md"><img src="https://agentmods.dev/badge/instructions/rmednitzer/relay-shell/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 | $0.00716 | $0.00716 |
| Opus 5 | $0.00358 | $0.00358 |
| Sonnet 5 | $0.00143 | $0.00143 |
| Haiku 4.5 | $0.00072 | $0.00072 |
Grade A, and why
relay-shell 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 — 88 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md
Claude-specific collaboration guide for relay-shell.
Objective
Deliver robust, auditable, high-capability shell/SSH tooling for MCP clients, with support for real-world privileged administration (including root and sudo) when operators intentionally choose that posture.
Core behavior expectations
- Prefer surgical changes over broad rewrites.
- Maintain compatibility with existing tools and response formats.
- Keep security controls explicit, not implicit.
- Preserve operational power; add safeguards and visibility instead of disabling features.
Required development loop
- Inspect relevant source + tests. For any change touching
policy,redaction,audit,patterns,broker, or theRelay.runbody, step 1 isdocs/runbook.md§2 (Audit) — those modules are the trust boundary. - Run baseline checks (
ruff check .,ruff format --check .,mypy,pytest -q). - Implement minimal safe improvements.
- Add/adjust tests for changed behavior.
- Re-run checks.
- Validate with review/security scanners before finalizing.
Coding guidance
Error handling
- Do not leak raw tracebacks to end users by default.
- Prefer structured, bounded error strings through existing error helpers.
- Keep exceptions contained inside tool execution wrappers.
Auditability
- Audit every tool call through the central runner.
- Keep redaction applied to audited arguments.
- Never write command output bodies to audit logs.
- Include meaningful metadata when safe (
request_id,client_id, exit code, tier).
Security posture
- Keep policy denylist first and global.
- Keep
open,guarded,readonlymode semantics stable. - Treat privileged execution cues (
sudo,doas,pkexec) as elevated-risk commands. - Do not remove SSH host verification controls; only make behavior more explicit.
Capability posture
The project intentionally supports:
- local shell execution
- SSH command/session execution
- file transfer and forwarding
- privileged admin flows (including root and sudo) when deployment allows
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 · 88 lines · 716 tokens per session scan A 09caa1d53804
relay-shell CLAUDE.md is an instructions file published in the GitHub repository rmednitzer/relay-shell (1 stars, last pushed 19d ago), licensed Apache-2.0. It adds 716 tokens to every session, about $0.0036 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
Containarium CLAUDE.md
Claude Code instructions for FootprintAI/Containarium, covering containarium — guidance for claude / ai assistants, no instance / endpoint / tenant names in oss-visible files, cli-first, mcp wraps it, api: protobuf contract first, grpc-gateway for rest and strong typing — use the type system, not strings.
vibeshell AGENTS.md
AGENTS.md instructions for veithly/vibeshell, covering vibeshell — agent guide, architecture, critical knowledge, russh 0.44 constraints and frontend → backend communication.
mcp-ssh CLAUDE.md
Instructions for AiondaDotCom/mcp-ssh, covering claude.md, project overview, development commands, basic operations and development scripts.
winkterm AGENTS.md
Instructions for Cznorth/winkterm, covering agents.md, project overview, common commands, backend development and start the development server.
Unfour AGENTS.md
AGENTS.md instructions for zyqzyq/Unfour, covering agents.md, product and module scope, package boundary rules, backend and command-bus rules and ai execution rules.
hop CLAUDE.md
Instructions for danmartuszewski/hop: Use conventional commit format with imperative mood.