Objective
Propose a portable decision-brief skill that reads one living work issue ledger and produces the next high-signal maintainer decision packet.
Why
We want one issue per unit of work. The machine should turn the current ledger plus trusted human amendments into the next bounded decision handoff without fragmenting context across separate proposal threads or hidden memory.
Desired Contract
- Input is portable subject memory, with a GitHub issue thread as one adapter surface.
- The skill reads the work issue ledger plus trusted amendment comments.
- It emits one bounded decision packet for the next human step: approval ask, blocker summary, open-question bundle, or maintainer handoff.
- It preserves receipts and keeps the issue thread as the human-visible ledger.
- It stops at an explicit handoff boundary and does not silently publish.
What We Want To Learn
- Whether skill-lab stays tight and useful on a small, legible problem.
- Whether the generated proposal distinguishes subject memory, outputs, receipts, and handoff boundaries cleanly.
- Whether a later amendment comment in this same issue sharpens the proposal instead of restating the whole thread.
Constraints
- Prefer one skill with a small, inspectable surface.
- No provider-locked nouns in the core contract.
- Keep the proposal grounded in the current runx / aster vocabulary.
Objective
Propose a portable decision-brief skill that reads one living work issue ledger and produces the next high-signal maintainer decision packet.
Why
We want one issue per unit of work. The machine should turn the current ledger plus trusted human amendments into the next bounded decision handoff without fragmenting context across separate proposal threads or hidden memory.
Desired Contract
What We Want To Learn
Constraints