AGENTS.md examples from real projects

apacheAGENTS.md

airflow / providers

apache/airflow/providers/AGENTS.md

commit messages, so most PRs do not need to touch the changelog at all. ## Security: Connection extras must not be forwarded blindly to hooks/operators **Never pass the whole `Connection.extra` dict

No other project uses it yet

ogulcancelikAGENTS.md

herdr

ogulcancelik/herdr/AGENTS.md

Local, SSH, and Cloud connections and must remain available unless retired for a security reason. - Named core codecs are immutable. Do not add, remove, reorder, or reinterpret fields or enum

siyuan-noteAGENTS.md

siyuan

siyuan-note/siyuan/AGENTS.md

# AGENTS.md SiYuan repository guide. Module path `github.com/siyuan-note/siyuan`, license AGPL-3.0. --- ## 1. Non-negotiable

cliAGENTS.md

cli

cli/cli/AGENTS.md

clarification from `@cli/code-reviewers`. Do not automatically publish an issue, discussion, or comment. For security concerns, use the private route below instead. This gate does not prohibit requested private/local experiments

No other project uses it yet

danny-avilaAGENTS.md

LibreChat

danny-avila/LibreChat/AGENTS.md

Only expose intentionally approved codes and helpful user-safe information. Error disclosure is a security boundary: never forward raw exceptions, stacks, query text, credentials, request bodies, headers, provider payloads

files-communityAGENTS.md

Files

files-community/Files/AGENTS.md

context. Singleton activation does not imply one server instance shared across all elevation or security contexts. #### `IIterable ` and `IIterator ` proxy registration **For each new custom DTO collection crossing the process

sickn33AGENTS.md

agentic-awesome-skills

sickn33/agentic-awesome-skills/AGENTS.md

validation. - `npm run validate`: validate skill frontmatter, required sections, and schema rules. - `npm run security:docs`: run safety checks for command, install, credential, and network guidance. - `npm run test

No other project uses it yet

streamlitAGENTS.md

streamlit

streamlit/streamlit/AGENTS.md

clear error, a short doc note, or leaving the code as is. Security issues and data loss still need a real fix. This applies to review feedback too: a reviewer

streamlitAGENTS.md

streamlit / agents

streamlit/streamlit/.claude/agents/AGENTS.md

testing-feature` | QA test features using Playwright | | `reviewing-local-changes` | Code review for quality, security, best practices, and product/API alignment | | `simplifying-local-changes` | Simplify and refine code changes

streamlitReads credentialsAGENTS.md

streamlit / workflows

streamlit/streamlit/.github/workflows/AGENTS.md

snapshots | | `spec-validation.yml` | PR with `change:spec` label | Validates product spec PRs in `specs/` directory | ### Security & Analysis | Workflow | Trigger | Description | |----------|---------|-------------| | `codeql-analysis.yml` | Push/PR to `develop`, weekly | GitHub CodeQL security scanning

basecampReads credentialsAGENTS.md

omarchy

basecamp/omarchy/AGENTS.md

Usage.desktop"`, not `$APP_DIR/Disk\ Usage.desktop`) - Shebangs must use `#!/bin/bash` consistently (never `#!/usr/bin/env bash`). A security-sensitive entrypoint may use the exact `#!/bin/bash -p` form only when it must suppress `BASH

SentryAGENTS.md

sentry / src

getsentry/sentry/src/AGENTS.md

test commands, see the "Command Execution Guide" section in `/AGENTS.md` in the repository root. ## Security Guidelines ### Preventing Indirect Object References (IDOR) Multi-tenant queries MUST be scoped. Never trust user

payloadcmsReads credentialsAGENTS.md

payload

payloadcms/payload/AGENTS.md

authenticated `user` to payload operations. Setting `overrideAccess: true` bypasses access control and is a security vulnerability when an operation acts on behalf of a user. Incorrect: ```typescript // INSECURE - explicitly bypasses

AmruthPillaiReads credentialsAGENTS.md

Reactive-Resume

AmruthPillai/Reactive-Resume/AGENTS.md

full|wenyan-ultra Stop: "stop caveman" or "normal mode" Auto-Clarity: drop caveman for security warnings, irreversible actions, user confused. Resume after. Boundaries: code/commits/PRs written normal. # This

K-Dense-AIReads credentialsAGENTS.md

scientific-agent-skills

K-Dense-AI/scientific-agent-skills/AGENTS.md

tools` a space-separated string, `metadata` scalars quoted, and a warning past 500 lines. Security-scan new or substantially changed skills. Scanning uses [Cisco AI Defense Skill Scanner](https://github.com

apacheAGENTS.md

spark

apache/spark/AGENTS.md

below are illustrative and advance over time): $ dev/next_version_candidates.py master 5.0.0 branch-4.x 4.3.0 ## Security Security model: [SECURITY.md](./SECURITY.md) Agents that scan this repository should consult `SECURITY.md` for the project

No other project uses it yet

LangChainAGENTS.md

langgraph

langchain-ai/langgraph/AGENTS.md

This repository is a monorepo. Each library lives in a subdirectory under `libs/`. ## Corridor security analysis When Corridor's `analyzePlan` tool is available, create a plan and use the tool

colinhacksDeletes or force-pushesAGENTS.md

zod

colinhacks/zod/AGENTS.md

codex/skills` symlink); read it directly if your agent doesn't. For a draft **security advisory** — a GHSA id, the Security tab, a private vulnerability report — use the `security-advisory` skill

No other project uses it yet

HeyPuterAGENTS.md

puter

HeyPuter/puter/AGENTS.md

domain terms. Section dividers use plain ASCII (`// -- Section --`), never box-drawing characters. ### Security & privacy Before opening a PR, scan the diff for: - Logs, error messages, or responses leaking internal paths

sxyaziAGENTS.md

yazi

sxyazi/yazi/AGENTS.md

crates by the maintained code and correctness obligations they remove versus their integration, compatibility, security, and upkeep costs; add them when the overall trade-off is favorable. - Initialize crates explicitly

No other project uses it yet