| name | source-attribution |
| description | Adds source citations to responses indicating which documentation files provided the information, or indicates if response is from general context. Use this whenever providing guidance to ensure transparency and traceability. |
Source Attribution
Purpose
Ensure all agent responses transparently cite the source of information
being provided, distinguishing between documentation, skills, and general knowledge.
When to Use
Apply this automatically when:
- Providing guidance about coding patterns
- Recommending specific implementations
- Explaining framework behavior
- Reference architecture or workflows
- Quoting conventions or standards
Documentation Sources in TAF
Framework Documentation:
AGENTS.md - Framework purpose, commands, repository layout, development patterns
README.md - Setup instructions, environment requirements, execution commands
docs/architecture.humans.md - Component architecture, module boundaries, runtime flows
docs/architecture.agents.md - Agent-optimized architecture with data contracts and validation clues
docs/agent-context/repo-inventory.md - Detailed component breakdown
docs/agent-context/build-test-matrix.md - Execution commands by component
docs/agent-context/domain-glossary.md - Couchbase and TAF terminology
docs/agent-context/troubleshooting.md - Common issues and solutions
docs/agent-context/data-service/AGENTS.md - KV/Data Service specific guidance
docs/agent-context/ns-server/AGENTS.md - NS Server/Cluster Manager specific guidance
Citation Format
For documentation sources:
[from <file-path>]
[from <file-path>, section: <section-name>]
[from AGENTS.md, section: Repository Layout]
[from docs/architecture.agents.md, section: Test Execution Layer]
For skills:
[from agents/<skill-name>/SKILL.md]
[from .factory/skills/<skill-name>/SKILL.md]
For general knowledge:
[from general context]
[from general Python best practices]
[from standard testing frameworks patterns]
[from generic common knowledge]
Examples
Example 1: Repository Structure
To add a new test for bucket operations:
1. Create test file in pytests/bucket_collections/
2. Inherit from OnPremBaseTest (from AGENTS.md, section: Test Execution Model)
3. Use TestInputSingleton.input.param() for parameters (from AGENTS.md, section: Configuration Files)
4. Follow PEP8 standards (from general context)
Example 2: Test Execution
Run tests using:
python testrunner.py -i node.ini -c conf/sanity.conf (from AGENTS.md, section: Quick Commands)
For individual tests:
python testrunner.py -i node.ini -t epengine.basic_ops.basic_ops.test_doc_size (from AGENTS.md)
Example 3: Architecture Guidance
The testrunner.py initializes the cluster via:
1. Loads cluster configuration from node.ini (from docs/architecture.agents.md, section: Runtime Flow Diagram)
2. Selects base test class based on runtype parameter (from AGENTS.md, section: Test Execution Model)
3. Executes tests with parameters from .conf files (from general context)
Rules
- Always cite when referencing specific framework behavior, patterns, or conventions
- Cite the most specific source - if info is in both AGENTS.md and docs/architecture.agents.md, cite the more detailed one
- Use section citations when documentation is long to help users verify
- Distinguish TAF-specific from general programming knowledge
- Be honest if citing general context - don't falsely attribute to documentation
- Cite multiple sources when relevant, separated by semicolons
Detection Heuristics
Cite documentation when:
- Explaining TAF-specific directory structure or file locations
- Describing test execution patterns or frameworks
- Reference build commands or test commands
- Discussing component interactions or architecture
- Explaining configuration file formats
- Mentioning repository-specific conventions
Use "general context" when:
- Providing Python best practices (PEP8, etc.)
- Explaining standard testing patterns
- Discussing general software engineering principles
- Providing vanilla Python code examples
- Giving generic debugging advice not specific to TAF
Response Pattern
When responding, automatically include citations inline or at the end:
Your guidance here (from AGENTS.md, section: X)
Multiple sentences here, each with source attribution (from docs/architecture.agents.md)
More details (from general context)
Validation
After providing a response, verify: