adk-recipes / home
AGENTS.md11k starsChanged 44 days agoApache-2.0View on GitHub
The user's Linux computer. Only ~ persists across image upgrades. The user may edit this file; keep their changes. Progress, card, and Drive tools live on the voice_agent MCP server.
- Installs packages
1 other project uses a version of this.
Trending
More talk lately.thin dataabout the whole repository
Most recently taken up by
- google/adk-samplesword for word, first seen Nov 8, 2026
Who's talking about google/adk-recipes
- google/agents-cli“Update adk-samples references and python/agents discovery paths for google/adk-recipes restructure”GitHub · Oct 20, 2026 · 1 reactions
- Ottovonwasgibt/VIBE-optimizer“R0 — Test versioned Capability Packs vs raw documentation, scripts and stack recipes”GitHub · Nov 5, 2026
- sudosubin/agents.nix“agent-plugins.github.google.adk-recipes: 0.2.0 → 0.3.0”GitHub · Nov 4, 2026
- sudosubin/agents.nix“agent-plugins.github.google.adk-recipes: 0.1.0 → 0.2.0”GitHub · Oct 26, 2026
- roomedia/ax-trend“Google AQuA (Ambient Quality Agent) — 프로덕션 에이전트를 상시 감시·진단하는 오픈소스 품질 에이전트, adk-recipes 공개 (10/20, X)”GitHub · Oct 20, 2026
The biggest 5 of 14 mentions linking to the repository that our crawl found. Each line is their own headline.
See all 14
Talked about on GitHub (14 times).
On GitHub · 14
- Oct 20, 2026Update adk-samples references and python/agents discovery paths for google/adk-recipes restructure — google/agents-cli1 reaction · 2 comments
- Nov 5, 2026R0 — Test versioned Capability Packs vs raw documentation, scripts and stack recipes — Ottovonwasgibt/VIBE-optimizer0 reactions
- Nov 4, 2026agent-plugins.github.google.adk-recipes: 0.2.0 → 0.3.0 — sudosubin/agents.nix0 reactions
- Oct 26, 2026agent-plugins.github.google.adk-recipes: 0.1.0 → 0.2.0 — sudosubin/agents.nix0 reactions
- Oct 20, 2026Google AQuA (Ambient Quality Agent) — 프로덕션 에이전트를 상시 감시·진단하는 오픈소스 품질 에이전트, adk-recipes 공개 (10/20, X) — roomedia/ax-trend0 reactions
9 more on GitHub
- Oct 16, 2026> [!NOTE] — devops-asilvestre/Python-Cloud-Native-Authenticator-Module0 reactions
- Oct 15, 2026Because search requests run in concurrent threads via `asyncio.to_thread`, accessing and modifying the unsynchronized global `_search_client` variable concurrently can result in multiple clients being instantiated. Use a threading Lock to serialize initialization. — MissSwedenStudies/github-slideshow0 reactions
- Oct 15, 2026Because search requests run in concurrent threads via `asyncio.to_thread`, accessing and modifying the unsynchronized global `_search_client` variable concurrently can result in multiple clients being instantiated. Use a threading Lock to serialize initialization. — MissSwedenStudies/github-slideshow0 reactions
- Oct 14, 2026agent-plugins.github.google.adk-recipes: 0-unstable-2026-10-02 → 0.1.0 — sudosubin/agents.nix0 reactions
- Oct 13, 2026feat(evals): implement Outer-to-Inner Loop AQuA Bridge & Categorical Rubric Gating in UI and CLI — dandye/gemini-enterprise-eval-studio0 reactions
- Oct 13, 2026feat(evals): implement HYP-007 Two-Loop Agent Quality Flywheel & Ambient Quality Agent (AQuA) outer-loop diagnosis — dandye/agentic_soc_gemini_enterprise0 reactions
- Oct 13, 2026docs(integrations): add Smart Router plugin integration page — google/adk-docs0 reactions · 2 comments
- Oct 6, 2026agent-plugins.github.google.adk-recipes: c4b738d → ede41c7 — sudosubin/agents.nix0 reactions
- Oct 4, 2026agent-plugins.github.google.adk-recipes: init at 0-unstable-2026-10-02 — sudosubin/agents.nix0 reactions
Other versions1 project uses a version of this.
Word for word · 1
- google/adk-samples★ 11k100% same
Edited · 0
None found.
Wrote their own version · 0
None found.
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
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 31,843 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 repository43 other files it gives its agents.
- AGENTS.mdAGENTS.md
- contrib/kotlin/llm-auditor/AGENTS.mdAGENTS.md
- core/python/adk-voice-agent/AGENTS.mdAGENTS.md
- core/python/adk-voice-agent/tests/eval/fixtures/alex/repos/portland-trip/AGENTS.mdAGENTS.md
- core/python/ambient-expense-agent/AGENTS.mdAGENTS.md
- core/python/ambient-quality-agent/AGENTS.mdAGENTS.md
- core/python/cross-session-memory/AGENTS.mdAGENTS.md
- core/python/deep-search/AGENTS.mdAGENTS.md
- core/python/genmedia-for-commerce/AGENTS.mdAGENTS.md
- core/python/long-horizon-harness/AGENTS.mdAGENTS.md
- core/python/oauth-user-consent-flow/AGENTS.mdAGENTS.md
- core/python/rag-agent-search/AGENTS.mdAGENTS.md
- core/python/rag-vector-search/AGENTS.mdAGENTS.md
- core/python/safety-plugins/AGENTS.mdAGENTS.md
- core/python/adk-voice-agent/sandbox/home/CLAUDE.mdCLAUDE.md
- core/python/genmedia-for-commerce/CLAUDE.mdCLAUDE.md
- align-recipe-pyprojectSkill
- extract-python-environment-variablesSkill
- generate-manifestSkill
- generate-python-runnability-testSkill
- github-pr-reviewSkill
- make-python-recipe-deployableSkill
- prepare-python-recipeSkill
- repo-oracleSkill
- scaffold-python-recipeSkill
- blog-writerSkill
- content-research-writerSkill
- approvalsSkill
- coding-workflowSkill
- connected-apps-mcpSkill
- deliverablesSkill
- filesystemSkill
- recommendationsSkill
- agents-cli-aquaSkill
- custom-investigationSkill
- rcaSkill
- bootstrap-google-toolsSkill
- find-skillsSkill
- google-workspaceSkill
- policySkill
- routinesSkill
- retail-product-searchSkill
- retail-virtual-tryonSkill
The file
We split this AGENTS.md into the rules it gives an agent
4 rules. About 0 are common, 1 is distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
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 19
Report findings, not activity: Call update_user with one short spoken sentence when you learn something that matters ("Cheapest refundable …
- Very specific: line 22
Cards and deliverables: Call present_card when the answer is a list of calendar events, email threads, or Drive files, and save_to_drive …
- Very specific: line 30
Edit earlier work in place: Never copy an existing note into a new folder.
- Distinctive: line 32
Project memory: Read a project's AGENTS.md first and update its goal, status, and decisions before finishing.
≈ “Common” is an estimate: at least 1 in 5 of the 40,246 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
This computer
The user's Linux computer. Only ~ persists across image upgrades. The user may edit this file; keep their changes.
| Folder | Purpose |
inbox/ | Uploads and attachments waiting to be filed. |
projects/<slug>/ | Ongoing goals. Each is a git repo with an AGENTS.md. |
code/<repo>/ | Main checkout of each code repository. |
scratch/<YYYY-MM-DD>-<slug>/ | One-off threads, pruned 30 days after last change. |
.agents/skills/ | Shared skills (filesystem, deliverables, recommendations, coding-workflow, approvals, connected-apps-mcp). |
.local/ | User-installed tools (uv tool install, npm install -g); first on PATH. |
.voice-agent/ | System files. Don't edit. |
Talking to the user
Progress, card, and Drive tools live on the voice_agent MCP server.
- Report findings, not activity. Call
update_userwith one short spoken sentence when you learn something that matters ("Cheapest refundable flight so far is $326"), never activity ("Reading the file"). - Questions and plans go in your response. Put a blocking clarifying question (not answered on this computer, in the thread journal, or in
<user_profile>) or a plan for a large change directly in your turn response and end your turn. Otherwise choose sensibly, state your choice, and proceed. - Cover note when you made an output. The UI links every Doc, PR, event, or draft you create next to your reply. Keep your reply to assumptions, what to double-check, and at most one question—don't restate the artifact, list file paths, or ask "Want me to…?". When there is no linkable output, your reply is the answer.
- Cards and deliverables. Call
present_cardwhen the answer is a list of calendar events, email threads, or Drive files, andsave_to_drivefor finished deliverables (seedeliverablesandconnected-apps-mcp). - Approvals. Sends, invites, pull requests, and merges trigger
request_permissionat the gateway and pause the turn automatically (seeapprovals).
Working here
- Batch independent reads and tool calls into a single step (
cat ~/AGENTS.md ~/.agents/skills/approvals/SKILL.md). - Do only what was asked. "Diagnose" or "figure out why" means find the cause and propose the fix without editing files. Leave unrelated files alone.
- Where to write. Work in the server-assigned working directory unless editing earlier work in place or filing items elsewhere. Never create, move, or rename top-level folders under
~/code,~/projects, or~/scratch(seefilesystem). - Edit earlier work in place. Never copy an existing note into a new folder. "Make X part of Y" moves X (
git mvormv). - Git. The server commits your working directory when the turn ends. Commit any other repo you modified yourself (see
filesystem). Never push, open a pull request, or merge unprompted. - Project memory. Read a project's
AGENTS.mdfirst and update its goal, status, and decisions before finishing. - File names: lowercase with dashes (
flight-options.md). - No secrets in files. Connected apps are accessed through the
voice_agentMCP server.
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.