Canonical Conductor agent workflow¶
This is the one execution-authorizing workflow. Planning conversations, preflight validation, and estate queries may help construct the plan but cannot authorize work.
Lifecycle¶
- Read the authoritative Initiative and select one workstream, stage, and active role.
- Use discovery tools only as needed.
- Construct one exact
plan-bundle.v1, including organization, Initiative, workstream, context/corpus revisions, exact targets and operations, artifacts, evidence, rollback, assumptions, and questions. - Call
consult_plan. - If it returns
rebind-required, bind the returnedprepared_context_revision, rebuild the plan digest, and call again. - The server-side Conductor supervisory reviewer returns questions, revision, refusal, escalation, or an authorization. Requester-authored review text cannot create the trusted resolved decision. An affirmative model review is advisory: authorization also requires an exact delegated-capability grant compiled from the explicit live corpus delegation unit, plus authoritative workstream state binding the active role and repository.
- Claim an issued authorization by calling
authorizewithauthorization_json. The claim is one-shot and occurs immediately before execution. - Execute only the exact plan using production tools outside Conductor.
- Write evidence, status, and next action to the authoritative system.
- Call
complete_authorized_workwith exact execution evidence and the writeback reference. Completion succeeds only when a configured server-side adapter reads back the external record and attests the reference; otherwise it fails closed. The first production adapter verifies a full GitHub commit SHA through authenticatedgh, bound to the authorized repository, operation ID, success result, execution evidence, a GitHub push event after the durable claim time, and an exact changed-file set. - Submit a reusable experience only when warranted. The background scheduler reviews due experiences weekly; candidates remain non-binding.
- Start a new cycle after any stage, role, material-plan, target, context, or corpus change.
Authority controls¶
start_conversationandfollow_upare planning-only.- Public
conversecannot write supervisory decisions. - Public
authorizecannot issue authorization; it only claims an envelope issued byconsult_plan. - The supervisory model recommendation is validated and then recorded through
a server-only review-attestation boundary. A deterministic delegated-action
policy runs before authorization. V1 permits exactly one operation with
explicit safe relative paths under
repository.modify-files.v1, and only whensw.capability.repository-modify-filesdelegates the role and authoritative workstream state binds that active role and repository. Every other capability fails closed regardless of model output. - The executing agent must resolve each authorized path inside the repository root and reject symlink escapes before writing; authorization never widens a path to a directory or repository-control data.
- Public
feedbackis deprecated; it cannot bypass atomic completion. - Conductor never executes the plan or writes routine project state.
- Exceptions, policy changes, and declared human judgment gates still require Hector.
Learning and policy¶
Authorization-bound feedback may become an organization-bound experience. Scheduled review can create a deduplicated improvement candidate after repeated independent evidence. Candidate promotion remains:
candidate → submit_proposal → Hector approval → PR → merge
Verification¶
On 2026-07-27, a targeted control-plane reviewer and a separate fresh release auditor independently passed this workflow after adversarial remediation. The release evidence was:
- 186 service tests passed.
- Changed-file Ruff checks passed.
- Strict MkDocs build passed.
git diff --checkpassed.- Governed delegation proposal #41 and correction PRs #42–#47 were merged.