← ClaudeAtlas

sprintlisted

Focused execution from a concrete outcome to SHIPPED, PAUSED, FAILED, or ABORTED_BY_USER. Adds acceptance-driven loops, bounded review, and delivery evidence beyond native coding behavior. Routes unclear requirements to grill, bugs to debug, UI work to ui, and completed delivery to ship. NOT for hypothetical or planning-only responses.
panda850819/verbs · ★ 5 · Code & Development · score 77
Install: claude install-skill panda850819/verbs
# Sprint A sprint owns one finish line and ends `SHIPPED`, `PAUSED`, `FAILED`, or `ABORTED_BY_USER`; local edits are never `SHIPPED`. **Planning-only boundary:** for a hypothetical, plan-only, or no-tools request, do not enter the state machine. Return `Execution: NOT_RUN` without inventing commands or evidence. ## 1. Bind the finish line Read the repository contract and the issue, brief, plan, or request. State one outcome, in-scope files or subsystem, checkable acceptance evidence, exclusions, and irreversible operations. If one un-derivable choice changes the result, route only that question to `grill`; route bugs to `debug` and UI work to `ui`. For an existing plan, treat it as read-only context; re-derive progress from git and acceptance checks instead of hand-editing status. Completion: the finish line can be proven by commands or named human evidence. ## 2. Execute tight loops For each smallest coherent unit: inspect the seam, make the minimum change, run the narrowest relevant check, and inspect the diff. Execute in the main session by default; delegate only file-disjoint or mechanical work requested by the user or repository contract through `handover`. Consult `lib/model-anchors.md` for the model seat; never select a model ad hoc. Stop immediately before destructive or public actions that need authorization. Never weaken or skip a test. Completion: every changed line maps to the finish line and its check passes. ## 3. Verify the artifact Run the real accep