Repository navigation
fix(agents): let Skald own ship delivery - #91
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Skald Code’s host prompts told coordinators to fix small review findings themselves and to watch a PR when a merge was pending. Both can redirect an active ship run away from the repository’s engine profile and Skald’s landing gate.
For ship requests, the shared instructions now defer staffing and PR creation to the repo skill, use Skald MCP for delivery, and report missing access instead of falling back to T3 or GitHub. T3 still links PRs for visibility, handles standalone forge requests, and watches a PR when asked. This changes guidance, not tool availability, permission checks, or merge behavior.
Added tool-use evals for worker assignment, a green-but-unmerged head, and a denied Skald grant; updated the standalone merge controls. Refreshed the Muse replay fixtures. The measured Claude and Codex context is 39 bytes smaller than the base branch, without a budget exception.
Validation:
vp run check:changed origin/atlipassed (546 tests, typechecks for 27 packages); all 8 Muse replay scenarios passed. Paired GPT-6.1 Sol tool-use eval: 7/12 samples passed on the base prompts, 12/12 on the new prompts (six simulated cases, two samples each). Both standalone merge-permission controls improved from 0/2 to 2/2; the missing-access case improved from 1/2 to 2/2. The worker, Publisher gate, and explicit watch controls passed on both. The worker grader checks actual dispatch and absence of coordinator edits because this harness can encrypt the child brief.Model: gpt-6.2-astra · Harness: Codex in Skald Code.