Appendices: working templates and glossary

The templates below were created for this report. Adapt them to the size and consequences of the task; fields that don't help a decision can be removed.

Appendix A — Task contract for agents

Problem and user:
  What need or failure should be resolved? For whom?

Expected result:
  What observable behavior should exist upon completion?

Reliable context:
  Which files, documents, contracts, and examples should guide the work?
  Which version or initial state is in use?

Invariants:
  What must remain true, including under failure?

Scope:
  What may change? Which decision or access requires another owner?

Evidence:
  How will behavior, integration, and consequences be verified?
  Which risks require specific cases?

Execution:
  What resources are available? What is the budget and stopping condition?
  How will progress be recorded and the task resumed?

Completion:
  Deliver the change, evidence tied to the version, limits, and open items.

Appendix B — Contents of a short repository map

The map should point to the details. Replicating all the content across several files creates more places to maintain and reconcile.

Appendix C — Failure and flow improvement log

Task and configuration:
Observed symptom:
Reproducible evidence:
Expected result:

Investigated cause:
  requirement / context / environment / tool / model / verification / other

Chosen intervention:
  minimal change that addresses the cause

Validation of the intervention:
  case that used to fail; current result; conditions verified

Shared knowledge:
  test, tool, document, or procedure changed

Follow-up:
  owner; recurrence signal; condition for revising or removing the rule

Appendix D — Evidence-driven review of a delivery

Use these questions to locate gaps, adjusting the depth to the change:

• Does the result solve the agreed problem?
• Are the domain decisions explicit?
• Does the change preserve the relevant contracts and invariants?
• Was the evidence produced on the version that will be merged or deployed?
• Does the verification exercise the important risks, including failures and concurrency where applicable?
• Are there relevant differences between the test environment and the environment of use?
• Do the access, dependencies, and external effects correspond to the authorized scope?
• Will the behavior be observable, and is there a recovery path?
• Does the final record distinguish verified facts, hypotheses, and open items?
• Should any discovery improve the next work cycle?

Appendix E — Brief glossary