feat(loop-init): scaffold loop-intake skill to clarify ambiguous goals before acting#237
Open
KhaiTrang1995 wants to merge 1 commit into
Open
feat(loop-init): scaffold loop-intake skill to clarify ambiguous goals before acting#237KhaiTrang1995 wants to merge 1 commit into
KhaiTrang1995 wants to merge 1 commit into
Conversation
…s before acting The design checklist asks for a "single clear goal" and primitives warn against guessing, but nothing helps a loop when the work item it receives is too vague to verify "done". Issue-driven loops in particular act on human-authored input that often lacks a testable definition of done, exact scope, or the concrete values a fix depends on -- so the loop guesses and burns fix attempts the circuit breaker then has to catch. This adds a cross-cutting loop-intake skill (same family as loop-guard / loop-constraints / loop-budget): - templates/SKILL.md.loop-intake: runs at the front of a run when the item is underspecified. Ask one question at a time, push for exact values, and on a still-vague answer write an open question + needs-human escalation instead of guessing. Report-only by default; feeds loop-verifier via a "Done when" line. - loop-init scaffolds it for issue-triage (PATTERNS_NEEDING_INTAKE), mirroring how the circuit breaker is scaffolded only for fix-capable patterns. Report-only patterns without ambiguous input skip it. Wired into README and the loop-design checklist. Two tests cover presence for issue-triage and absence for daily-triage. Version 1.3.3 -> 1.4.0. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Contributor
|
Thanks @KhaiTrang1995 for contributing a docs improvement — visible, reviewable PRs like this grow the reference for everyone. What happens next
More ways to help — loop-engineering maintainers |
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.
Summary
The design checklist asks for a "single clear goal" and the primitives warn
against guessing, but nothing helps a loop when the work item it receives is too
vague to verify "done". Issue-driven loops in particular act on human-authored
input that often lacks a testable definition of done, exact scope, or the
concrete values a fix depends on -- so the loop guesses and burns fix attempts
the circuit breaker then has to catch. This adds the missing front-door.
What
A cross-cutting
loop-intakeskill, in the same family as loop-guard /loop-constraints / loop-budget:
is underspecified. Ask one question at a time, push for exact values, and on a
still-vague answer write an open question +
needs-humanescalation instead ofguessing. Report-only by default; feeds
loop-verifiervia a "Done when" line;respects
loop-constraintsscope.issue-triage(PATTERNS_NEEDING_INTAKE),mirroring how the circuit breaker is scaffolded only for fix-capable patterns.
Report-only patterns without ambiguous input skip it.
Why this shape
four tools and is off-domain (intake is not a scheduled loop). A skill template
is the right unit;
bundle-assetspicks it up automatically at build.norm) to keep special-casing minimal and the behavior easy to test.
Tested
npm testgreen (14/14). Two new tests: loop-intake is scaffolded forissue-triage and is absent for report-only daily-triage. loop-init README and the
loop-design checklist updated; version 1.3.3 -> 1.4.0.