[codex] complete governed repair demo workflow - #1
Merged
Merged
Conversation
Keep a reproducible 16K Ollama configuration for offline AgentTeams demos on the development laptop.
Require approved Skill metadata before publication and bind grant issuance to the pinned Skill version, role, tool, path, risk, and approval context. Align task terminal and platform-blocked states with the architecture.
Pin five addyosmani/agent-skills sources by commit and content hash, define strict role and permission metadata, and add input/output schemas without claiming evaluation or publication.
Add explicit investigation, approval, independent verification, rollback, learning, and terminal transitions. Keep the authoritative architecture state diagram aligned with the executable workflow.
Records immutable v1.1.2 source and image identities, plus the four-role AgentLoom team and scoped Human resource without credentials.
Validates pinned images, applies resources in dependency order, waits for Matrix room membership, and emits redacted deployment evidence.
Adds environment-only Qwen provider activation and strict Matrix sender-owned E2E evidence so the four-role runtime can be validated without leaking credentials.
Stages immutable Manager parent tasks into team storage and collects only allowlisted result artifacts after hash verification, closing the v1.1.2 global/team handoff gap without modifying upstream images.
Adds a secret-free Docker mc adapter, CLI evidence output, and documented stage/collect workflow. Live smoke validation confirmed allowlisted artifacts cross namespaces while workspace files remain isolated.
Adds typed investigator, implementer, verifier, and risk outputs with cross-artifact task and patch hash invariants required by the deterministic demo pipeline.
Adds a synthetic Python defect that must fail before repair, a fixed patch replay, independent test/static verification, evidence hashing, state transitions, and contract-valid role artifacts without invoking an LLM.
Publishes a deterministic parent task, stages it into team storage, collects only allowlisted role artifacts, and revalidates the bundle from global MinIO. The live smoke run uses no LLM and stays explicitly labelled Mock.
Run frozen cases through bounded commands, target-specific reproduction, isolated hidden verification, and patch-scope enforcement. Migrate the original fixture and prove the same artifact flow with a second defect type.
Expose manifest-defined cases, four-role task status, workflow events, and verified repair artifacts without a cloud model dependency.
L2 grants now require an exact, approved, unexpired request record; L3 execution is blocked for the competition runtime. The SQLite ledger prevents stale decisions and invalid approval rows.
The internal control plane now creates, reads, and version-checks approval decisions with structured conflict and not-found responses.
WilliamClifton-dev
changed the title
WilliamClifton-dev
marked this pull request as ready for review
August 4, 2026 14:55
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.
What changed
Why
AgentLoom needs a reproducible, evidence-first software repair demonstration with distinct AgentTeams roles, independent verification, explicit approval boundaries, rollback behavior, and transparent third-party attribution.
Validation