ARCKnowledge-Android
AGENTS.md1 starsChanged 19 days agoMITView on GitHub
Subagents are autonomous executors — they take a task, use tools, and return a result. This is distinct from Skills, which are knowledge containers that load patterns for Claude to apply. These agents exist (definitions in ARCAgentStack config/agents/, symlinked into ~/.claude/agents) and are the ones CLAUDE.md routes to. ARC agents come from ARCAgentStack (config/agents/), symlinked into ~/.claude/agents — nothing to install per project. ARC Android skills live in this repository's .claude/skills/ and are symlinked into each project by ARCDevTools-Android's arc-setup…
Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.
Of the AGENTS.md files we read, the 33,315 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.
In this repository5 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- android-doctrineSkill
- android-testing-doctrineSkill
- android-verifySkill
- arc-cli-androidSkill
The file
We split this AGENTS.md into the rules it gives an agent
5 rules. About 0 are common, 2 are distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 25
Minimum privilege: read-only agents (arc-verify-android, arc-audit-material) have no Edit/Write tools
- Distinctive: line 28
No commit/push: all agents with write access are constrained from committing or pushing
- Very specific: line 40
Create config/agents/arc-[name].md in ARCAgentStack with frontmatter (name, description, model, tools), using the arc-agent-authoring skill
- Distinctive: line 42
Add entry to this AGENTS.md (Active Agents) and to the agents table in CLAUDE.md
- Very specific: line 43
Add entry to Skills/skills-index.md under the Agents section
≈ “Common” is an estimate: at least 1 in 5 of the 42,283 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
ARC Labs Android Subagents
Subagents are autonomous executors — they take a task, use tools, and return a result. This is distinct from Skills, which are knowledge containers that load patterns for Claude to apply.
| Concept | Role | How to trigger |
| Skills | Load reference patterns — Claude applies them | android-doctrine, android-testing-doctrine, etc. (see Skills/skills-index.md) |
| Agents | Execute autonomously — Claude delegates the task | "Implement X with TDD", "Review before commit", "Debug this error" |
Active Agents
These agents exist (definitions in ARCAgentStack config/agents/, symlinked into ~/.claude/agents) and are the ones CLAUDE.md routes to.
| Agent | Model | Read-only | Role |
arc-verify-android | Sonnet | Yes (Bash, Read, Grep, Glob) | External verifier — zero-warning Gradle build, unit tests, Compose preview screenshot validation, install-and-walk on a headless emulator via the android CLI. Iterate until it returns a pass. |
arc-audit-material | Sonnet | Yes (Bash, Read, Grep, Glob) | Audits UI work against Material 3 Expressive, the Navigation 3 shell, edge-to-edge, adaptive layout and TalkBack. One blocker and the UI work unit does not close. |
Design Principles
- Minimum privilege: read-only agents (
arc-verify-android,arc-audit-material) have no Edit/Write tools - Dynamic skill invocation: agents invoke skills relevant to the task, not all skills by default
- Model selection: the agent's frontmatter pins its model; both active agents run on Sonnet
- No commit/push: all agents with write access are constrained from committing or pushing
Installation
ARC agents come from ARCAgentStack (config/agents/), symlinked into ~/.claude/agents — nothing to install per project. ARC Android skills live in this repository's .claude/skills/ and are symlinked into each project by ARCDevTools-Android's arc-setup (scripts/setup-skills.sh).
Adding a New Agent
- Create
config/agents/arc-[name].mdin ARCAgentStack with frontmatter (name,description,model,tools), using thearc-agent-authoringskill - List the skills it will invoke dynamically in its instructions
- Add entry to this
AGENTS.md(Active Agents) and to the agents table in CLAUDE.md - Add entry to
Skills/skills-index.mdunder the Agents section
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

