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 skills/sharpninja/mcpserver/bash-usagenpx skills add sharpninja/McpServer --skill bash-usagegit clone --depth 1 https://github.com/sharpninja/McpServerWhat 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 | $0.00072 | $0.01515 |
| Opus 5 | $0.00036 | $0.00758 |
| Sonnet 5 | $0.00014 | $0.00303 |
| Haiku 4.5 | $0.00007 | $0.00152 |
Grade A, and why
bash-usage 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 yesterday.
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 — 104 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Bash Usage on the McpServer Windows Host
This skill governs shell selection for the optional QBAgent (McpServer.QBAgent, Quad-Brain)
run_bash and run_powershell tools. The host OS is Windows 11. PowerShell 7+ (pwsh.exe)
is the primary and always-present shell. Git Bash is optional and only usable when it is
resolvable on PATH.
When to Use
- A QBAgent tool loop needs to run a shell command and you must choose between
run_powershellandrun_bash. - A
run_bashcall returnedavailable: falseand you need the correct fallback. - You are porting a POSIX shell snippet (the build doc examples use
./build.ps1 ...but show bash fences) and need to decide which tool actually executes it.
When Not to Use
- Pure file read/write/search work. Use the dedicated file tools, not a shell.
- Session log, TODO, requirements, import/export, or traceability operations. Those must
go through the
mcpserver-claude-code-pluginwrapper (workflow.*/client.*methods), not throughrun_bashorrun_powershell. Nevercat,grep, orsedthedocs/Project/TODO.yamlor session-log files from a shell. - Long-lived servers or builds you intend to background. Prefer the project's documented
entry points (
./build.ps1 Compile,./build.ps1 Test,./build.ps1 StartServer).
Inputs
- The shell command string you intend to run.
- The QBAgent tool result fields, specifically the
availableboolean returned byrun_bash. Whenavailableisfalse, Git Bash was not found onPATHand no command was executed. - Whether the command is POSIX-specific (uses
/dev/null,&&chains over POSIX builtins,grep/sed/awk, single-quoted here-docs) or is already PowerShell-native.
Critical Rules
run_powershell(pwsh) is the primary shell on this host and is always available. Default to it. The repo standard ispwsh.exe(PowerShell 7+); never targetpowershell.exe(Windows PowerShell 5.1).run_bashis OPTIONAL. It only works when Git Bash is onPATH. If the tool returnsavailable: false, treat bash as unavailable for this run and do not retry the same call expecting a different result.- Do not invent shell tool names, flags, or fields. Only
run_powershellandrun_bashexist for QBAgent shell execution, and the only availability signal is theavailablefield on therun_bashresult. - When
run_bashis unavailable, translate the command to PowerShell and run it throughrun_powershell. Do not assume a bash path or callbash.exedirectly. - Use absolute paths in commands. The agent working directory is not guaranteed to persist between tool calls.
- PowerShell syntax differs from POSIX: use
$nullnot/dev/null,$env:VARnot$VAR, backtick for line continuation, andpwsh-native cmdlets instead ofgrep/sed/cat.
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.
- yesterday First seen · 104 lines · 0 tokens per session scan A df311599a30f
bash-usage is a skill published in the GitHub repository sharpninja/McpServer (2 stars, last pushed 11d ago), licensed Apache-2.0. It adds 72 tokens to every session and 1,515 once invoked, about $0.0004 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 skills, from other repositories
opencli-sitemap-author
Use when creating or maintaining OpenCLI site sitemaps: agent-facing navigation, page-state, action, workflow, API-reference, pitfall, and fallback knowledge for a website. Use after browser exploration discovers durable site context, when a sitemap is stale, or when promoting local site knowledge into the repo.
golden-rss
Use when testing the rss golden build.
omh-code-review
This is a Hermes-native code-review workflow skill.
redteam-web-detail-pack
Routing and boundary guidance for authorized general web application security testing. Use as a web testing router when the attack surface should be dispatched to more specific web vulnerability skills.
android-pentest
安卓应用渗透测试 — APK分析、Hook、自动化测试、运行态驱动、签名恢复、抓包分析.
studio
Architecture Studio control plane — initialize or inspect a studio workspace, create and register projects, or route an architecture/AEC task to the right agent or skill. Use when the user runs /as:studio, asks to set up or open their studio, manage its projects, or describes a task without naming a skill.