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/thewolffish/wolffish-app/systemnpx skills add thewolffish/wolffish-app --skill systemgit clone --depth 1 https://github.com/thewolffish/wolffish-appWrote 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/skills/thewolffish/wolffish-app/system)<a href="https://agentmods.dev/skills/thewolffish/wolffish-app/system"><img src="https://agentmods.dev/badge/skills/thewolffish/wolffish-app/system.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 | $0.00029 | $0.01856 |
| Opus 5 | $0.00015 | $0.00928 |
| Sonnet 5 | $0.00006 | $0.00371 |
| Haiku 4.5 | $0.00003 | $0.00186 |
Grade A, and why
system 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 — 174 lines — stays where its author put it; the contents beside it link to each section on GitHub.
System & application control
Open and close apps, open files/folders/URLs in their default handler, and control the machine's power state. These tools issue native OS commands, so they work without the browser or computer-use automation.
When to use
- The user names an app to open or close ("open Spotify", "close Chrome").
- The user wants to open a file, folder, or link with its default app ("open my Downloads folder", "open this PDF", "open github.com").
- The user asks to restart, shut down, sleep, lock, or log out the machine.
- You need to know what apps are currently open before acting.
Tools
app_open— launch an app by name; optionally open a file/URL in it.app_quit— quit an app by name (graceful by default;forceto kill it).app_list— list the currently open GUI apps.open_path— open a file/folder/URL with the OS default handler;revealshows a file in the file manager instead of opening it.system_power—restart·shutdown·sleep·lock·logout;delaySecondsschedules a restart/shutdown instead of firing it now.
Rules
- Match the user's intent exactly. "Close X" →
app_quit(graceful). Only passforce: trueif they explicitly say force/kill or a graceful quit already failed — it can lose unsaved work. restart,shutdown, andlogoutrequire confirmation and will be shown to the user for approval before running. Don't call them speculatively; only when the user clearly asked.sleepandlockrun without a prompt.- Never take the machine down at zero delay. A power action is the one tool
call whose effect outlives the turn that made it: when the machine goes, this
turn is still being written — the answer, the tool cards, the task timeline,
and on a cloud device the copy the org has not received yet.
delaySecondsdefaults to 20 for exactly that reason. Leave it alone unless you have a reason to raise it. Pass0only if the user explicitly said to go down immediately, and tell them what it costs. - Reboot last, and say so. Make the power call the FINAL action of the turn
— never mid-plan with steps still queued behind it. Before calling it, finish
and report the work, then tell the user in one line what happens next: that
the machine restarts in N seconds, what it will apply, and (on Windows) that
shutdown /acancels. A user who learns about the reboot from the screen going black was not warned. - Give the machine longer when something is still running. Raise
delaySecondsif a long write, a download, an upload, or a background job you started is still in flight — the delay is the only window those have to finish. Seconds are cheap; a half-written file is not. - Don't reboot through the shell.
shutdown /r,Restart-Computer,reboot, andosascript ... to restartviashell_execskip the approval card AND the delay, which is the whole safety net. Usesystem_power. - Prefer
open_pathover the shell. To open a file/folder/URL, useopen_path, notshell_execwithopen/xdg-open/start. - Use the real app name. On macOS that's the display name ("Visual Studio
Code", not "code"). If a quit fails, the name may be wrong —
app_listshows what's actually open. - Connected channels are not apps. Telegram and WhatsApp are messaging
channels Wolffish connects to, not local apps — never
app_open,app_quit, orapp_listthem (or osascript a "Telegram"/"WhatsApp" window) to send or read messages. Use the channel tools (telegram_send,whatsapp_send, …) instead.app_openis only for launching unrelated local apps the user explicitly named. - These commands target the local machine the user is sitting at.
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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 Changed · +25 lines a067e5176370
- 6d ago First seen · 149 lines · 29 tokens per session scan A a75ab61e9dcd
system is a skill published in the GitHub repository thewolffish/wolffish-app (5 stars, last pushed yesterday), licensed MIT. It adds 29 tokens to every session and 1,856 once invoked, about $0.0001 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
agent-computer-use
REQUIRED for any task that involves operating a desktop application — opening apps, clicking buttons, typing into fields, pressing keys, scrolling, dragging, reading what's on screen, moving or resizing windows, or verifying state after an action. Always use the agent-cu CLI commands (open, snapshot, click, type, key…
claude-api
Build, debug, and optimize Claude API / Anthropic SDK apps. Apps built with this skill should include prompt caching. Also handles migrating existing Claude API code between Claude model versions (4.5 → 4.6, 4.6 → 4.7, retired-model replacements). TRIGGER when: code imports anthropic/@anthropic-ai/sdk; user asks for…
create-skill
Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy.
gui-integration-test
GUI desktop app only. Writes, runs, and debugs Warp integration tests using the custom Builder/TestStep framework in crates/integration. Use when adding a new integration test, fixing a failing integration test, wiring a test into the manual runner or nextest suite, or verifying end-to-end UI and terminal behavior in…
gui-onboarding-verification-skill
GUI desktop app only. Launch two parallel Oz cloud agents with computer use to download and install the latest stable Linux Warp build, capture screenshots while walking through first-time onboarding in both logged-out and logged-in states, then selectively fan out follow-up cloud agents for distinct onboarding…
logging-and-error-reporting
How and when to log (log:: levels, safe macros) and report errors to Sentry (reporterror!) in the Warp codebase. Use when adding or reviewing any logging or error reporting — picking a log level, deciding log vs. reporterror!, keeping sensitive data out of logs, or surfacing an error to Sentry.