ROS Agent Bootstrap

Follow this sequence before beginning work.

1. Read operating instructions

1. framework/REP-SPECIFICATION.md
2. framework/policies/RESEARCH-POLICY.md
3. framework/policies/EVIDENCE-POLICY.md
4. framework/policies/OUTPUT-POLICY.md
5. framework/protocols/ARTIFACT-LIFECYCLE.md
6. framework/protocols/SUPERSESSION.md
7. Any standards or discipline-specific profiles named by the mission

2. Read current shared context

1. context/CURRENT-STATE.md
2. context/ARCHITECTURE.md
3. context/DECISIONS.md
4. context/KNOWN-RISKS.md
5. Relevant project or domain context

3. Read the assigned mission

Read only the mission and source records needed to execute it. Do not load the entire repository without a reason.

4. Execute

• Identify the largest material uncertainty.
• Form explicit hypotheses.
• Seek supporting and contradicting evidence.
• Attempt to falsify conclusions.
• Preserve source traceability.
• Use deterministic tools before generative inference where practical.
• Record failures and invalidated assumptions.

5. Record results

Create or update the appropriate canonical artifacts:

• Journal: research/journals/
• REP: research/packages/
• Evidence: research/evidence/
• Hypotheses: research/hypotheses/
• Theories: research/theories/
• Experiments: research/experiments/

Run ./ros registry build; registries are generated and must not be edited manually. Then run ./ros validate.

6. Handoff

The next agent must be able to reconstruct the work, understand the current theory, and continue without additional conversational context.