FO smallest-sufficient-mechanism gate — stop spurious orchestration / busywork#478
Merged
Merged
Conversation
Resident blockquote in first-officer-shared-core Working Principles, directly adjacent to z25's evidence-bar bullet (the both-directions gate + scope clause); the ladder, named-busywork list, and worked examples ride the new lazy reference fo-smallest-sufficient-mechanism.md. Resident delta +996 bytes. AC-1 groundwork per the four-step recipe: the smallest-sufficient-mechanism shared runtime scenario (deterministic in-house edits + convention-direct commit, no worker/PR climb) plus the commissioned-engage scope guard (the gate stays silent through the standing dispatch loop), with the mandatory offline negative grading the mechanism trace red on each incident. Live drives are validation's. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…engage dialect Cycle-1 validation found the codex GRADER (not the gate or scenario) false-negatived a correct live run: codexMechanismTrace read in-house edits only from apply_patch command_executions, but codex-cli 0.142.5 applies edits as native file_change items, and the real standing-dispatch engage surfaces as `status --set <entity> status=done` (the `wait` collab carries no per-entity binding), not spawn_agent. Extend the extractor to read structured item types: file_change changes[].path for in-house edits, `status --set … status=done` for engage, and agent_message text for the per-entity gate justification — critically NOT command strings, since the FO's contract-reads contain "smallest sufficient mechanism" (a false-positive source; 15 command hits vs 0 agent_message hits on the correct run). Offline negatives cover the new surfaces incl. the contract-read false-positive guard. Fresh codex-live drive greens (356s; file_change edits + status-set advances + zero PR). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Contributor
Journey cost delta
|
Collaborator
Author
|
Merging with codex-live red under explicit captain waiver (2026-07-07): NOT the known spawn-oracle flap this time — shallow-boot failed because the driven codex FO fumbled state-sweep call shapes (tried 4 wrong flag forms, never 'state sweep --workflow-dir .') and improvised a partial manual advancement; transcript-verified NOT caused by this diff's mechanism gate (the FO actively pursued the sweep; zero gate vocabulary near the decision). Recorded in seed task sc (live-runner-boot-preamble-hardening). Merge bar met: claude-live GREEN on both variants (sonnet on rerun after a boot-preamble flake, also recorded in sc; the Claude extractor's first live exercise), pi-live green, all deterministic lanes green. Two validation cycles on record. |
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.
Stops the FO from manufacturing orchestration for work it can do directly — a smallest-sufficient-mechanism gate on discretionary mechanism choices, with commissioned standing dispatches explicitly exempt.
What changed
Evidence
Review guidance
The codex extractor rework (item-types over command regexes) is the risk center — cycle-1/2 in the audit link.
zm