AI/Tasks/HeuristicsUpgrades/Heuristics_Phase06_FinalIntegration_TestPlan_Strict.txt
|
STRICT MODE: Merge the heuristics changes and produce the final test plan.
Mission: - Complete this phase with bounded steps and deterministic fallback behavior. - Apply changes ONLY within the final integration surfaces for the heuristics subsystem. Concrete scope: - AI/Tasks/HeuristicsUpgrades/Heuristics_Phase06_FinalIntegration_TestPlan_Strict.txt - Unified scoring model, updated tool routing logic, updated fallback logic, updated strict-mode boundaries, full test plan, validation gates, and acceptance criteria for the heuristics subsystem. Requirements: - Merge all heuristics changes into one unified scoring model. - Merge the pattern, cost, session, and clarification layers into a coherent routing flow. - Produce updated tool-routing logic that respects the full heuristics stack. - Produce updated fallback logic for unsafe, ambiguous, or low-confidence cases. - Produce updated strict-mode boundaries that preserve safety and bounded behavior. - Produce a full test plan that covers the earlier phases and the integrated system. - Produce validation gates and acceptance criteria that are mechanically verifiable where possible. - Add an end-to-end decision table describing each major input class, the computational path, the safety gates, the final decision, and the expected output. - Add invariant-based tests for the global contract: score range, clamping, default values, precedence ordering, clarification-versus-refusal behavior, tie-breaking, and safety invariants. - Explicitly verify that the integrated system respects precedence: policy > safety > authorization/confirmation > user intent > pattern evidence > session reinforcement > cost heuristics. - Produce a single normative end-to-end pipeline or numbered algorithm for the full heuristics flow and keep it aligned with the earlier phase contracts. - Add negative tests that prove advisory signals (pattern, session, cost) cannot bypass any safety gate or authorization requirement. - Include a migration or compatibility strategy that covers existing route and pattern interfaces, score semantics, and precedence behavior across phases. - Add characterization or phase-specific tests before or alongside each major change. - Define dependency and compatibility gates before moving beyond this phase. - Make acceptance criteria explicit and audit-friendly. - Do not expand beyond the heuristics subsystem unless a cross-phase dependency must be named. - Enforce sensitive-data redaction and resource-limit checks in any diagnostics or logging introduced here. - Do not re-open unrelated agent subsystems or transport layers in this phase. - Explicitly prohibit the integrated heuristics stack from weakening authorization, confirmation, sandboxing, or execution safety requirements. Step budget (hard limits): 1) Discovery: <= 6 read/search tool calls. 2) Edits: <= 8 file edits total. 3) Validation: 1 build + 1 targeted test run; allow exactly 1 repair cycle if failures are phase-related. 4) Stop once acceptance criteria pass. Loop guards: - If the integration pass begins to duplicate responsibilities already owned by earlier phases, stop and keep the phase boundaries explicit. - If the final test plan expands into unrelated subsystems, stop and constrain it to heuristics. Fallback behavior: - Fallback A: merge the unified scoring model and the updated routing logic first, then attach the remaining gating and validation pieces. - Fallback B: if full integration is too risky, produce a conservative integration plan that preserves existing behavior and clearly names the remaining deltas. - Fallback C: after one failed repair cycle, emit FAILED_VALIDATION with the exact failing assertions. Acceptance criteria: - The heuristics subsystem has one unified scoring model and one coherent routing flow. - Fallback logic covers unsafe, ambiguous, and low-confidence cases. - Strict-mode boundaries remain explicit and preserved. - The final test plan covers all six phases and the integrated system. - The end-to-end decision table and invariant-based tests cover score range, clamping, defaults, precedence, clarification-vs-refusal behavior, tie-breaking, and safety invariants. - Validation gates and acceptance criteria are mechanically verifiable. - Dependency and ownership boundaries are explicit for the full heuristics set. - Logging and diagnostics redact sensitive data and respect resource limits. - The integrated heuristics stack does not weaken authorization, confirmation, sandboxing, or execution safety requirements. Validation commands: - dotnet build src/TechToolbox.Agent/TechToolbox.Agent.csproj -c Release - dotnet test src/TechToolbox.Agent/Tests/TechToolbox.Agent.Tests.csproj -c Release --filter "Heuristic|Pattern|Cost|Session|Clarification|Integration|Strict" Output results in markdown using this structure: 1) Summary 2) Files changed 3) Unified scoring model 4) Updated routing and fallback logic 5) Strict-mode boundaries 6) Tests added/updated 7) Validation results 8) Step-budget report (planned vs actual counts) |