forsonny

Use this agent during the contracting skill's Process step 3 to draft `anvil/contract.yml` in a fresh context. Pass the user's source_intent, the extracted pattern tags, and the Ledger query results. The agent writes the file, validates it, and returns ONE sentence. Do not invoke it for anything else.

Use this agent during the planning skill's Process to decompose the confirmed `anvil/contract.yml` into an atomic task DAG at `anvil/plan.yml`, in a fresh context. Pass the repository path. The agent writes the file, validates it, and returns ONE sentence.

Stop the current run. Captures a reason and, when applicable, a final lesson.

Resume from saved state. Reads anvil/state.json and picks up at the next incomplete task.

Query, append, or audit the global lesson Ledger.

Finalize a passing run. Open a PR via gh pr create. The human reviews and merges; ship does not merge for the user.

Begin a new task stream. Authors a contract from user intent via the contracting skill.

Meta-skill governing every change to any skill file. Requires a RED-then-GREEN subagent pressure transcript before a skill change is accepted.

Author a machine-readable contract before execution. Produces anvil/contract.yml with four verification-level slots per criterion and Ledger-sourced counter-examples.

Dispatch one fresh subagent per plan task inside a per-task git worktree. Capture diff and tool output; never carry state across dispatches.

Dispatch the Court on evidence-only inputs (contract + diff + verify output + confidence). The Court is not a persona; it is a structured adjudicator.

Decompose a confirmed contract into an atomic task DAG with wave ordering. Every task cites one or more contract criterion ids.

On verify or judge fail, append a non-null lesson to the Ledger and re-queue the task. Null lessons are refused; the task escalates instead.

Bootstrap loader for the DG-Anvil plugin. Enumerates the five primitives, seven skills, five commands, and the core loop. Loaded at session-start.

Run the four verification probes (Exists, Substantive, Wired, Functional) per criterion. All four green per criterion or the criterion fails.