Clarify orchestrator output budget and precondition checks - #47
Merged
Conversation
…ation The launch-ralph supervisor and the launch-implement front door told the model what to be rigorous about but never how little to say, so it proved compliance by narrating every precondition check and restating scope/target/engine several times — including a close-out-style recap minutes after launch. Add an explicit output budget: two reports per run (a one-line pre-launch echo and the close-out recap), everything else done silently unless it fails or changes the plan. Verify preconditions silently. Clarify in the Rules that target/engine are restated once at close-out, never in between. Prose-only tightening of the shipped skills; no policy, merge-gate, or workflow-logic change (ralph.workflow.js log() lines are already terse and left untouched). Build and full test suite (178) pass. Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Ez3MNLZDxUCcsiHZsxfUxk
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
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Refine the
launch-ralphandlaunch-implementskill documentation to establish clearer expectations around when the orchestrator should report status versus work silently.Summary
This change tightens the contract for the orchestrator's communication pattern, emphasizing that status updates should be minimal and strategic rather than continuous. It introduces the concept of an "output budget" — two required reports per run (pre-launch echo and close-out recap) — and specifies that all intermediate checks and state management should happen silently unless they fail or change the plan.
Key changes
Output budget policy: Explicitly define that the orchestrator owes exactly two reports per run:
Precondition verification: Add guidance that preconditions should be verified silently — a green base and reachable tracker are the expected case, not news. Surface a precondition only when it fails.
Echo as sole pre-launch report: Clarify in
launch-implementthat the echo is the one pre-launch report, and intermediate checks (scope resolution, setup repair, routing) should not be narrated unless they fail or change the scope.Target and engine declaration: Update the rules to specify that the integration target and engine are declared in the pre-launch echo and restated once in the close-out recap — never in between.
Implementation details
These changes reinforce the principle that the orchestrator is a decision-maker and supervisor, not a play-by-play narrator. By establishing a strict output budget, the skill documentation now makes clear that silence is the default state, and communication is reserved for moments that matter: the initial scope confirmation and the final outcome.
https://claude.ai/code/session_01Ez3MNLZDxUCcsiHZsxfUxk
Need help on this PR? Tag
@codesmith-botwith what you need. Autofix is disabled.