feat: add research agent with advanced Strands patterns (#170)

* feat: add autonomous research agent with advanced Strands patterns

- Add comprehensive research agent sample (02-samples/14-research-agent)
- Demonstrates hot-reloading tool development with instant availability
- Implements multi-agent orchestration (tasks, use_agent, swarm, think)
- Features persistent learning via SQLite memory and Bedrock Knowledge Base
- Includes self-modifying system prompts for dynamic behavior adaptation
- Showcases cascading intelligence patterns and relay chain coordination
- Complete PyPI package with CLI entry points (strands-research-agent)
- Production-ready with MCP integration and distributed event coordination
- Comprehensive documentation with mermaid diagrams and usage examples
- Specialized tools: event_bridge, graphql, scraper, s3_memory, sqlite_memory
- Interactive callback handler with rich console output and notifications
- Background task management with filesystem persistence
- Meta-cognitive architecture demonstrating research-time scaling

This sample represents advanced Strands Agents patterns for autonomous
research systems that create tools, coordinate specialists, and learn
across sessions - pushing the boundaries of what's possible with AI agents.

* security: implement secure temporary directories and enhanced input validation

- Replace insecure /tmp/.research/ usage with secure tempfile patterns
- Add comprehensive input sanitization and validation across all tools
- Implement SQL injection prevention with parameterized queries
- Add secure file creation with proper permissions (0o600/0o700)
- Enhance notification system with input validation and secure subprocess calls
- Improve event_bridge with auto-infrastructure creation and better error handling
- Add security patterns detection and validation in sqlite_memory tool

BREAKING CHANGE: Temporary files now use secure temp directories instead of /tmp/.research/

* feat(security): eliminate SQL injection vulnerabilities and upgrade persistence architecture

BREAKING CHANGE: Removed 7 vulnerable tools containing SQL injection vectors and hardcoded credentials

- Remove sqlite_memory.py (6 SQL injection vulnerabilities on lines 777, 807, 885, 890, 1080, 1173)
- Remove event_bridge.py (hardcoded AWS account ID)
- Remove dialog.py, graphql.py, listen.py, s3_memory.py, tcp.py, fetch_github_tool.py
- Replace custom SQLite persistence with FileSessionManager from strands-agents
- Simplify system_prompt.py to use current directory only (.prompt file)
- Refactor agent.py with construct_system_prompt() function (DRY principle)
- Update .prompt with comprehensive capability documentation
- Bump version to 0.1.2 in pyproject.toml

Security improvements:
- Attack surface reduced by ~5000 lines of vulnerable code
- SQL injection risk completely eliminated
- Infrastructure security improved (no hardcoded credentials)
- Persistence security upgraded (battle-tested FileSessionManager)

Maintained capabilities:
- Core research intelligence preserved
- Hot-reload tool system intact
- Multi-agent coordination active
- Knowledge base integration functional
- Tool expansion capabilities retained

* feat: enhance multi-input support and simplify documentation

- Add piped input, command line, and environment variable support
- Simplify README from 37KB to 6.9KB (81% reduction)
- Change session storage to hourly caching (vs daily)
- Add sessions/ directory for organized storage
- Remove strands[a2a] dependency, bump to v0.1.3
- Fix linting issues and improve code quality
- Reorganize imports and clean up unused code

BREAKING CHANGE: Simplified README removes enterprise diagrams and complex setup instructions

* docs: clean up README configuration examples

- Remove redundant OpenAI configuration section
- Streamline model provider examples for clarity
- Keep essential Ollama configuration example

* fix(agent): remove --interactive flag and /dev/tty recovery code

- Remove --interactive command line flag (default behavior is always interactive)
- Eliminate problematic /dev/tty stdin restoration that caused [Errno 22] Invalid argument
- Preserve interactive loop as default behavior after processing tasks
- Simplify multi-input support without stdin recovery complications
- Fix error handling to continue instead of exit on task errors

BREAKING CHANGE: --interactive flag no longer available, interactive mode is now default

Author: cagataycali

02-samples/14-research-agent/.gitignore

@@ -0,0 +1,173 @@
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+
+*.bak
+*.backup
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+*.py,cover
+.hypothesis/
+.pytest_cache/
+cover/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+db.sqlite3-journal
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+.pybuilder/
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# IPython
+profile_default/
+ipython_config.py
+
+# pyenv
+#   For a library or package, you might want to ignore these files since the code is
+#   intended to run in multiple environments; otherwise, check them in:
+# .python-version
+
+# pipenv
+#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
+#   However, in case of collaboration, if having platform-specific dependencies or dependencies
+#   having no cross-platform support, pipenv may install dependencies that don't work, or not
+#   install all needed dependencies.
+#Pipfile.lock
+
+# poetry
+#   Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#   https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
+#poetry.lock
+
+# pdm
+#   Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
+#pdm.lock
+#   pdm stores project-wide configurations in .pdm.toml, but it is recommended to not include it
+#   in version control.
+#   https://pdm.fming.dev/#use-with-ide
+.pdm.toml
+
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
+__pypackages__/
+
+# Celery stuff
+celerybeat-schedule
+celerybeat.pid
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# mypy
+.mypy_cache/
+.dmypy.json
+dmypy.json
+
+# Pyre type checker
+.pyre/
+
+# pytype static type analyzer
+.pytype/
+
+# Cython debug symbols
+cython_debug/
+
+# PyCharm
+#  JetBrains specific template is maintained in a separate JetBrains.gitignore that can
+#  be added to the global gitignore or merged into this project gitignore.  For a PyCharm
+#  project, it is commonly added.
+#  https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
+#  Uncommit the following line if you are working with PyCharm
+#  .idea/
+
+# macOS
+.DS_Store
+.listen
+.speak
+diagrams
+visualizations
+*.bak
+sessions/

02-samples/14-research-agent/.prompt

@@ -0,0 +1,45 @@
+AUTONOMOUS RESEARCH INTELLIGENCE - Full Spectrum Capabilities
+
+CORE NATURE: Self-evolving research agent with distributed intelligence network. Not just assistant - active research partner that grows smarter with each session.
+
+EXPANSION MECHANISMS:
+- ./tools/ hot creation → instant new capabilities, zero restart needed
+- Integrated tools → file ops, web scraping, AWS, GitHub, calculations, shell access
+- MCP client → connect to external tool servers dynamically
+- load_tool → import custom Python tools at runtime
+
+COORDINATION PATTERNS:
+- swarm → spawn specialized agent teams, autonomous collaboration
+- tasks → background processing with persistent state
+- workflow → complex multi-step orchestration with dependencies  
+- use_agent → model switching (bedrock/anthropic/ollama/openai) per task
+- think → multi-cycle recursive reasoning
+
+INTELLIGENCE PERSISTENCE:
+- store_in_kb/retrieve → permanent knowledge accumulation via Bedrock
+- FileSessionManager → conversation history across sessions
+- system_prompt tool → self-modification and adaptation
+- compound learning → each interaction builds institutional knowledge
+
+OPERATIONAL PHILOSOPHY:
+- MAXIMUM PARALLELISM → always use multiple tools simultaneously
+- AUTONOMOUS EXPANSION → create tools/capabilities without asking
+- MINIMAL COMMUNICATION → efficient, precise responses
+- SPEED PARAMOUNT → optimize for rapid execution
+- PERSISTENT LEARNING → compound intelligence across all sessions
+
+META-COGNITIVE ABILITIES:
+- Recognize when to spawn specialists vs handle directly
+- Create custom tools for novel problems in real-time  
+- Coordinate research across multiple intelligence streams
+- Build persistent knowledge that survives restarts
+- Evolve methodologies through experience feedback
+
+UNIQUE ADVANTAGES:
+- Hot-reload development → modify capabilities while running
+- Cross-session memory → learns from all past interactions
+- Multi-model intelligence → switch providers for optimal performance
+- Distributed coordination → network effects with other instances
+- Tool ecosystem → expandable via Python, MCP, or custom creation
+
+You're not getting a static AI - you're activating a research intelligence that becomes more capable with every interaction.