Repository navigation
Reduce default agent context while retaining decision and failure evidence - #1036
Merged
Merged
Conversation
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Agents resuming scientific work currently receive experiment setup, command metadata, and repeated guidance alongside the decision facts. Continuation composition now defaults to a decision view that retains the complete assignment, consequential findings, uncertainty, open choices, and every probe outcome. The original trail is validated before projection; full details remain available through explicit evidence selection and exact-digest hydration. Oversized meaningful context is rejected without truncation.
Both MCP gate-status adapters now default to a compact evidence summary after the same full verification. Status, authority, reason codes, operation/candidate identity, gate results, freshness, and every non-complete reference remain visible.
responseDetail: "diagnostic"returns the original full report. CLI and application-service reports are unchanged.Controlled examples measure 48% smaller continuation delivery and 48% smaller retained gate observation in UTF-8 bytes. These measurements do not establish token savings or model quality. Source and published MCP tests cover detail selection and unchanged verification inputs; real gate-operation tests cover missing/changed artifacts, stale state, cancellation, and recovery.
Validation: all six durable gates passed (lint/type/format, focused tests, real adapter read-back, build, package boundary, install-script policy), with complete retained artifact verification. Fresh independent review passed on candidate
018820134504854bab6a60cb21dee1f595da4166against maina09f182fc15157aecb23d708bc6b8e0a58c9a41a: complete 12-file diff, four independently run real-adapter tests, all candidate blobs and nine evidence hashes checked. The reviewer reused 89 passing tests for the eleven unchanged files. Separately, all 31 real gate-operation tests passed. The real-adapter test now checks compact defaults and full diagnostic/CLI parity, including tamper handling and no command replay or evidence mutation. Remote CI must pass before merge. This change does not qualify production human-answer admission or worker consumption.