[1.0 T3] Assemble bounded design context in Bash and PowerShell #179

Problem and outcome

Retrieve useful reasoning automatically while avoiding repeated project-wide reads and user-managed context selection.

Interaction contract

Keep FluencyLoop. Proposed positioning: Stay fluent as people and agents evolve the product.

Use retained product reasoning automatically. Explain consequential choices at the right depth. Recommend a path. Continue within agreed intent. Ask when human judgment is needed. Update the shared model at natural boundaries.

The flagship goal/principles/precedents/tradeoffs reasoning is integrated into the existing design explanation at proportional depth. It is not a mandatory recital, proposal form, or new approval exchange. People must remain able to understand and challenge the design.

Scope

• Use T1 evidence and the T2 resolver to select bounded context from explicit references, areas, relationships, and tags. Let the agent expand when needed without asking the person to manage retrieval.
• Reuse context within a task until relevant intent, files including uncommitted edits, principles, or decisions change. HEAD alone is insufficient for invalidation.
• Return revisions, selection reasons, truncation, gaps, and ambiguity as machine-readable context. Historical omissions are advisory and must not generate recurring user questions.
• Implement Bash/PowerShell parity with existing helpers; no Node requirement, database, vector search, or semantic completeness claim.

Acceptance criteria

• A seeded relevant precedent is available before design; replaced history is excluded from normal active context.
• Repeated unchanged slices reuse context instead of scanning/reloading the entire model.
• Relevant uncommitted changes refresh context even with unchanged HEAD.
• Missing brief or historical link does not block an authorized task when intent is sufficient.
• Parity fixtures cover limits, expansion, refresh, legacy data, and ambiguous references.

Validation

Add Bats/Pester selection and invalidation fixtures and record initial/unchanged/changed-input context sizes. Test without Node. Observe that omissions do not become questions unless they affect a real design choice.

Starting points

plugins/fluencyloop/fluencyloop, fluencyloop.ps1; scripts/{bash,powershell}/common., check.; tests/node-optional.bats; site/server.js resolution contract

Constraints

• Target dev; main distributes releases. No runtime implementation occurs in the planning PR.
• T1 is an observed pilot using existing artifacts before new infrastructure. Use its evidence to retain, narrow, or defer later scope.
• Recommend and continue within authorized intent; ask with a recommendation only for consequential unresolved judgment. Batch independent questions, reuse settled answers, and wait for dependent work. Silence is not approval.
• Absent briefs, historical link gaps, stale prose, and unreviewed metadata alone are advisory. Do not require onboarding, cleanup, or retrospective approvals to work.
• Unknown familiarity is not fluency and does not automatically require a blocking probe/check. Support requested guided learning without quizzes; keep calibration/preferences private. T5 explicitly updates inherited pause rules.
• Capture rationale once, link mechanical views, record outcome deviations at natural closure, distill once at wrap-up, and reuse unchanged context.
• Preserve append-only history/provenance; Bash and PowerShell implement the same contract. Core remains Node-optional; no new database, hosted service, or vector search.

Dependencies and size

• Task: T3
• Relative size: L (not a calendar estimate)
• Depends on: T2: [1.0 T2] Add stable record references and explicit decision lifecycle #178
• Milestone: https://github.com/baokhang83/fluencyloop/milestone/3
• Plan: draft source
• Planning PR: docs: propose FluencyLoop 1.0 shared product direction and design #189