This repository was archived by the owner on Oct 16, 2026. It is now read-only.
Repository navigation
Fix/execution - #190
Merged
Merged
Fix/execution#190
Conversation
- Extract FILE_CREATE from raw payload before sanitize, tolerate marker lengths - Resolve SEARCH/REPLACE creation body, skip anchor check when no original - Gate creations with verifier, record Created vs Changed, allow empty files - Truncation via finish_reason=length; short complete files accepted - Add create integrity and mission acceptance tests
- executor exempts create_file from constrained patch-only gate and disables hidden reasoning - selector binds prose-named absent creation targets to targeted_mutation - openrouter preserves explicit reasoning disable for non-whitelisted models - recovery re-prompts creation schema violations under same create contract - harness adds provider override plus exact-prompt live benchmark and creation tests
- normalize quote delimiters in classifier and strategy target extraction - bind ScopeDynamic for plain prompts inside /build as execution contract - carry command mode from per-run SetScope with subcommand fallback - add quoted-create, command-axis, build-contract and benchmark tests
- route /build ordinary prompt into authoritative Driver runtime with $build surface - add mutation_operation auth scope consumed once at operation terminal state - gate markup repair by target type and reject placeholder-echo artifacts explicitly - derive create proposal diff from artifact bytes, fail empty payload explicitly - track executionSurface for $prompt/$build/$hot and share $build dynamic authority - add ownership, integrity, verification and ingestion policy convergence tests
Subordinate the `izen prompt`/`izen orchestrate` proposal lifecycle (cli.Stack + orchestrator.RunCycle + ProposalProviderCLI + ProposalValidator + ApprovalGate + FileExecutor) onto the authoritative autonomy Driver over the RuntimeExecutor, reached through the production composition root (compose.Wire). - internal/cli/authoritative.go: narrow AuthoritativeDriver/HumanBridge ports + RunAuthoritative + terminal [y/N]/inspect bridge; renders the runtime's own CandidateDiff and issues the candidate-bound authorization only on approve. - cmd/izen/orchestrate.go: drive compose.Wire -> app.Autonomous; grant the workspace capability vector at admission; session-start checkpoint; preflight dispatch; audit flush; project runtime terminal truth. - objective_completion: map $orchestrate -> ScopeDynamic (same mutation authority as $prompt/$build; distinct command surface). - tests: behavioral ownership, legacy-vs-authoritative execution parity, AST seam/purity locks, $orchestrate provenance. - report: docs/report/IZEN_EXECUTION_TOPOLOGY_CONVERGENCE_PHASE2.md Live: Ollama qwen2.5-coder:7b CREATE completed with objective PROVEN by evidence in 2 provider calls; exact bytes applied; unrelated file intact. go test ./... exit 0 (212 ok, 0 FAIL).
- Resolve FILE_CREATE and contract fence to canonical bytes at staging before digest/preview - Derive CREATE diff from canonical bytes so approval preview matches applied file - Write canonical patches verbatim, fail closed on unterminated envelopes - Allow terminated empty FILE_CREATE as legitimate empty-file creation - Add integrity test and report
- run/orchestrate share runAuthoritativeCommand via compose.Wire and Driver - retire app.Pipeline and ledger from CLI path, delete runtime_ledger - executor resolves canonical bytes once for preview, digest and apply - patch Resolved writes verbatim with validator-only probe - conversation intent drops observation, keeps response obligation - add integrity regressions for malformed patch and existing create - lock convergence with architecture ownership tests and reports
- run/prompt/orchestrate share authoritative exec with approve/json/timeout flags - add resume/runs plus report with documented exit codes and quiet mode - add models/auth/stats/config surfaces backed by durable ledger - gate approvals via ask/always/never bridge and checkpointed auth - propagate ctx to contract fence parse and fix CLI help/tests - add subprocess CLI benchmark harness, bench-cli target, report
- derive project-level scope from workspace evidence - mark broad objectives unverified on skipped gate - expose strategy operation for scope reasoning - persist human decisions as durable ledger events - surface derived-scope rationale in review - add orchestration regressions and truth report
- Driver selects interaction via proposal advisor, parks on adaptive proposal boundary - Accepted decisions compile into existing execution contract without direct auth - Rendered-layout verification satisfies broad objectives via browser evidence - UI renders adaptive proposal card with select, submit, and resume surface - Durable PROPOSAL_SNAPSHOT preserves revision and decisions for resume - Compose wires default advisor, rendered policy, checks, and verifier
- Restore parked proposal from ledger on rerun instead of forming competing revision - Add free-form adaptive answer editor with focus, validation, and submit path - Harden proposal Restore to refuse zero revision and empty strategies - Retain custom audit events in durable store with replay support - Stabilize R5 proven-progress test and expand multiphase coverage
- Match preflight guard to bounded patch dispatch for constrained models - Defer PROVEN commit until after downgrade gates with staged grant - Publish adequacy policy with rejected reasons and empty tally - Surface rendered opt-in skip and blocked verification summary - Commit autonomous usage deltas once per run to fix 2x billing - Add terminal lifecycle and constrained contract coverage plus report
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.