grill-ai

Run a lightweight plan -> act -> analyze experiment to answer a question the cheapest way that still gives a trustworthy verdict.

Discuss and select the smallest executable feedback element for a development goal, then hand it to to-loop to write one cycle without speculating about…

Automatically dispatch ready, independent Wayfinder tasks to subagents in parallel, continuing batch by batch until no safe work can proceed.

Find the next actionable Wayfinder node; advance implementation according to its recorded approach, or review its implementation evidence against the recorded…

Run d-handoff, then compile today's journal, research notes, and handoff into a report with a table of contents and an Introduction/Abstraction/Detailed…

Run one selected NN-slug cycle as a single feedback loop; capture the expected/current gap, change the target code, and rerun to verify the change.

Design all approved requirement slices as small object-oriented vertical-slice solutions, then write one approved design-and-contract document per slice.

Run a @skills/grill-me interview to refactor a named function or usecase sequence against references/abstraction-levels.md's L1/L2/L3 rule — current shape and…

Execute a @skills/to-plan document's action items in order, checking each off in place, validate it, review the diff before commit, commit only after human…

Actively gather and capture durable session intent, facts, and constraints through clarification, research, and user decisions.

Personal grilling engine — interview the user round by round over a design-tree frontier; any question the user can't answer gets progressive-disclosure…

Explore the existing codebase, then turn prototype and repository context into an approved inventory of vertical requirement slices with concrete edge cases…