AI/Tasks/HeuristicsUpgrades/Heuristics_Phase05_ClarificationModeUpgrade_Strict.txt
|
STRICT MODE: Upgrade the clarification mode for heuristics.
Mission: - Complete this phase with bounded steps and deterministic fallback behavior. - Apply changes ONLY within the clarification layer and its direct interfaces. Concrete scope: - AI/Tasks/HeuristicsUpgrades/Heuristics_Phase05_ClarificationModeUpgrade_Strict.txt - Vague detection, ambiguity detection, missing target detection, missing verb detection, missing constraints detection, clarification templates, clarification triggers, and clarification overrides. Requirements: - Refine vague-detection rules so unclear requests are caught early. - Refine ambiguity detection so conflicting intent routes to clarification instead of unsafe action. - Refine missing-target detection so the system asks for a concrete target when needed. - Refine missing-verb detection so the system asks for an action when the intent is underspecified. - Refine missing-constraints detection so the system asks for the minimum necessary constraints. - Define reusable clarification templates that are concise and specific. - Define clarification triggers that are deterministic and easy to test. - Define clarification overrides for explicit user or system intent that should bypass a question. - Explicitly define the boundary between clarification and refusal: clarification handles underspecified, ambiguous, or missing required context; refusal handles unsafe, unauthorized, forbidden, or policy-violating requests. - Define a clarification loop limit and safe stop condition to prevent repeated or circular questioning; if the same missing fact remains unresolved after the allowed loop budget, stop and escalate to a conservative refusal or fallback. - Define the default clarification round limit, what counts as the same unresolved fact, and the exact fallback after the loop budget is exhausted. - Define a precedence order for multiple missing facts and a deterministic template-selection table for the most appropriate question. - Define safe fallback behavior for repeated unresolved ambiguity and preserve already supplied context without re-asking known facts. - 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 mechanically verifiable where possible. - Do not expand into the pattern library, cost multipliers, or reinforcement policy unless required to define clarification ownership. - Enforce sensitive-data redaction and resource-limit checks in any diagnostics or logging introduced here. - Do not change unrelated tool selection or safety policy in this phase. - Explicitly prohibit clarification logic 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 clarification rules begin to produce repetitive or redundant prompts, stop and collapse them into the smallest useful template set. - If override logic becomes broader than clarification ownership, stop and narrow the scope. Fallback behavior: - Fallback A: implement only the highest-risk missing-target and ambiguity checks first. - Fallback B: if full template coverage is risky, keep one conservative clarification template and one override path. - Fallback C: after one failed repair cycle, emit FAILED_VALIDATION with the exact failing assertions. Acceptance criteria: - Vague, ambiguous, and underspecified cases trigger clarification deterministically. - Clarification templates are concise and reusable. - Clarification overrides are explicit and safe. - Characterization or phase-specific tests cover the clarification behavior. - Dependency and ownership boundaries are explicit for any shared heuristics work. - Logging and diagnostics redact sensitive data and respect resource limits. 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 "Clarification|Ambiguity|Missing|Vague|Override" Output results in markdown using this structure: 1) Summary 2) Files changed 3) Clarification triggers 4) Clarification templates 5) Clarification overrides 6) Tests added/updated 7) Validation results 8) Step-budget report (planned vs actual counts) |