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 skills add felinics/Autofish --skill autofish-controlgit clone --depth 1 https://github.com/felinics/AutofishWrote 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/felinics/autofish/autofish-control)<a href="https://agentmods.dev/skills/felinics/autofish/autofish-control"><img src="https://agentmods.dev/badge/skills/felinics/autofish/autofish-control.svg" alt="Measured on agentmods" height="20"></a>- NVIDIA SkillSpector pass
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 | $0.00042 | $0.01687 |
| Opus 5 | $0.00021 | $0.00843 |
| Sonnet 5 | $0.00008 | $0.00337 |
| Haiku 4.5 | $0.00004 | $0.00169 |
Grade A, and why
autofish-control scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 8d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 165 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Autofish Control Skill
Use deterministic control with evidence at every step.
When To Use
- You need to navigate Android UI through Autofish (
af). - You need stable replay using refs (
@nK) instead of coordinates. - You need explicit verification after each action.
Setup
af config set remote.url "http://<IP>:<PORT>"
af config set remote.token "<token>"
af config set memory.db "$HOME/.config/af/af.db"
af config set output.default "text"
af config set artifacts.dir "$HOME/.config/af/artifacts"
Copy IP, PORT, and token from the Autofish Android app. For USB port forwarding, use the forwarded local address, for example http://127.0.0.1:<PORT>.
If the device is connected with adb and the App is new enough to write a connection hint, prefer:
af connect usb --device <ADB_SERIAL>
af connect usb writes the forwarded remote.url and connection.* USB metadata, verifies af health, and does not write remote.token. Keep the token copied from the app for observe, act, verify, and recover.
Environment variables also work: AF_URL, AF_TOKEN, AF_DB, AF_OUTPUT, AF_ARTIFACT_DIR.
af healthonly requires URL.observe,act,verify,recoverrequire URL + token.af memory ...is local-only and requires memory enabled.config listandconfig get remote.tokenredact tokens as<redacted>.- Use
--output jsonwhen parsing results in scripts. - Check
af --helpfirst if the installed CLI may be older than this skill. - If config/env is unavailable, put remote flags after the command group:
af --session <session> observe --url <url> --token <token> page.
Choose one session per task:
SESSION="task-name"
af --session "$SESSION" observe page --field screen --field refs --max-rows 80
Do not reuse the default default session across unrelated tasks.
Workflow
Memory session rules:
memory contextandmemory saveuse the global--session.memory logandmemory statsfilter with--for-session <session>.memory searchandmemory experiencequery cross-session history; do not treat global--sessionas a filter for them.
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.
- 8d ago First seen · 165 lines · 42 tokens per session scan A ba49827be122
autofish-control is a skill published in the GitHub repository felinics/Autofish (113 stars, last pushed 10d ago), licensed Apache-2.0. It adds 42 tokens to every session and 1,687 once invoked, about $0.0002 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 2026-08-30.
Other skills, from other repositories
open-gui-remote-control
Control an Android phone through OpenGUI from an AI coding agent such as Claude Code, Codex, or OpenCode. Use when an agent should list online devices, run a natural-language mobile task, check execution status, pause, resume, or cancel through the local OpenGUI backend and CLI.
open-gui-bootstrap
Launch and bootstrap OpenGUI from a plain-language user request. Use when an AI coding agent such as Claude Code, Codex, or OpenCode should install, configure, debug, and run the backend and Android client shipped in this repository while keeping manual setup to a minimum.
control
Automatically control locally connected Android phones with OpenGUI screenshots and allowlisted actions whenever a request targets a phone, Android app, mobile game, or multiple devices. The user does not need to name OpenGUI. Use the native Codex Browser instead for browser-only work.
earllm-build
Build, maintain, and extend the EarLLM One Android project — a Kotlin/Compose app that connects Bluetooth earbuds to an LLM via voice pipeline.
mobilerun-docs
Mobilerun documentation reference. Use when users ask about Mobilerun setup, configuration, SDK usage, CLI commands, device setup, agents, architecture, app cards, credentials, tracing, Docker, migration, structured output, or any Mobilerun "how do I..." questions.
screen-ocr-operator
Use when you need to see and operate an Android screen through an OCR/vision model plus ADB. Optimized commander workflow: one model plan, batch ADB execution, minimal round-trips, verify at milestones.