Research: bounded AgentOps approval workflow and synthetic control evaluation - #36
Draft
CodeWithJuber wants to merge 1 commit into
Draft
Research: bounded AgentOps approval workflow and synthetic control evaluation#36CodeWithJuber wants to merge 1 commit into
CodeWithJuber wants to merge 1 commit into
Conversation
AI-assisted research implementation; synthetic results only. Existing platform approval flow remains unchanged.
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 join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Mizan's current permission manager records pending approval but does not create a resumable execution grant. This research addition reuses its tool classification and implements a separate durable request lifecycle for a deliberately restricted synthetic incident application.
The new CLI supports propose, inspect, approve/reject, execute, and audit inspection across separate processes. Approvals bind exact plans, evidence, environment and policy. Policy checks, transactionally finalized requests, outcome observation and synthetic rollback support controlled comparisons.
Validation: 23 new tests pass; 96 episodes across 24 finite designed scenarios; a second run matches all semantic outcomes. Existing tests pass (557 passed, 8 skipped). Ruff passes.
make checkstops at the existing mypy duplicate-module configuration error for_version/backend._version; no existing backend source is changed.Scope: symbolic planner and SQLite simulator only, simulated operator decisions, no live LLM calls, real infrastructure adapter, bilingual evaluation, production benefit or general safety claim. Main Mizan chat/API approval behavior is unchanged. Forgekit and Hikmah inform the design; their runtimes are not integrated here.
Raw traces, source/dependency manifests, limitations and a proposed Hostlelo isolated staging pilot are included. This AI-assisted research artifact is not an IAAI submission. Keep this PR in draft while the remaining research and manuscript-authorship work is assessed.