claw/docs/superpowers/plans/2026-04-19-g6-real-sample-entry-preparation-and-bounded-execution-plan.md

G6 Real-Sample Entry Preparation And Bounded Execution Plan

Date: 2026-04-19 Status: Draft Upstream Spec: 2026-04-19-g6-real-sample-entry-preparation-and-bounded-execution-design.md

Plan Intent

This plan is the only surviving G6 execution plan after the redesign.

Its purpose is:

stop G6 planning recursion and move directly to one bounded implementation-plus-real-sample slice

Scope Guardrails

1. do not open any new G6 semantic sub-plan
2. do not reopen G7
3. do not open G8
4. do not open G4 / G5
5. do not broaden into host-runtime platform redesign
6. do not add more than one fixed G6 real sample

Preserved G6 Gate

The final frozen G6 gate under this plan is:

Hard Conditions

1. host-bridge-action-invocation-defined
2. callback-request-completion-defined
3. callback-state-verification-targets-defined

Soft Later Conditions

1. host-runtime-transport-implementation
2. real-sample-execution-proof

Fail-Close Reasons

1. g6_bridge_invocation_semantics_missing
2. g6_callback_completion_semantics_missing
3. g6_callback_state_targets_missing

Workstreams

1. WS1 Freeze the Final G6 Entry Gate
2. WS2 Implement the Minimum Host-Bridge Execution Seam
3. WS3 Run the Fixed G6 Real Sample
4. WS4 Write Back Validation And Close

Phase 0: Freeze The Final Gate

Objective

Stop semantic drift and declare the gate final for this execution slice.

Tasks

1. treat the hard G6 gate as frozen
2. treat the fail-close reasons as frozen
3. explicitly forbid any further G6 semantic micro-plan under this line

Deliverables

1. final frozen G6 gate note
2. final fixed-sample statement

Acceptance Criteria

1. no further G6 semantic clarification plan is produced

Phase 1: Implement The Minimum Execution Seam

Objective

Add only the minimum implementation needed to let the fixed G6 real sample enter one controlled execution attempt.

Tasks

1. implement the minimum host-bridge invocation seam required by the fixed sample
2. implement the minimum callback completion handling required by the fixed sample
3. keep the change narrower than generic host-runtime redesign
4. preserve fail-close behavior when the frozen hard conditions are not met

Deliverables

1. bounded G6 code change
2. bounded regression tests

Acceptance Criteria

1. G6 execution support is improved only at the seam required by the fixed sample
2. unrelated families are untouched
3. fail-close remains explicit

Phase 2: Execute The Fixed Real Sample

Objective

Use one real G6 sample to prove whether the bounded implementation slice is enough.

Tasks

1. run the fixed G6 real sample once
2. classify the result only as:
   • executed-pass
   • named mismatch
3. do not open a new semantic sub-plan regardless of result

Deliverables

1. real execution result
2. fixed-sample execution note

Acceptance Criteria

1. the result is narrower than “not executed”
2. the result is not deferred into another semantic-planning loop

Phase 3: Validation Closure

Objective

Write the fixed result back and close the line.

Tasks

1. update validation-layer assets
2. if pass: close G6
3. if mismatch: write one implementation correction plan only
4. publish a closure report

Deliverables

1. validation asset update
2. closure report
3. optional implementation correction plan if mismatch occurs

Acceptance Criteria

1. G6 ends in executed-pass or named mismatch
2. no new semantic micro-plan is emitted

Completion Criteria

This plan is complete when:

1. one bounded implementation seam is landed
2. one fixed G6 real sample is executed
3. the line closes with executed-pass or named mismatch

Non-Negotiable Stop Rule

After this plan starts executing:

1. do not create another G6 semantic plan
2. if the run fails, create only one implementation correction plan
3. if the run passes, close the G6 line immediately