Craft Goal

Craft Goal

Craft or lint the autonomy contract above AgentOps RPI. A goal is a persistent controller (the Goal / Mayor role) over a bead-shaped experiment graph: each RPI is one scientific trial, and the goal picks the next useful trial, preserves what was learned and ratchets toward a larger outcome.

Goal / Mayor: observe graph → choose bounded wave → consume results → ratchet
  └─ Bead: durable experiment intent, context, scratch, evidence, and links
       └─ RPI: plan → implement → checks → one fresh validate where a mistake is costly → report
            └─ Implementation: one RED → GREEN → refactor experiment

The number of RPIs need not be known in advance. A goal is safe when success is decidable, every experiment is bounded, evidence keeps its provenance, and the authorization envelope cannot silently renew itself. Beliefs are revisable: new evidence may retract an earlier claim. More stored knowledge is neither progress nor proof that knowledge is correct.

Authority boundary. The emitted prompt and safety report are inert caller-owned text. Crafting creates no goal, starts no runtime, mutates no bead and confers no standing authorization. The prompt drives RPI dispatch only when a caller pastes it into their own goal runtime, under their own authority and within the non-renewing envelope they set. Craft Goal reads tracker state when present but needs no tracker installed to compile a prompt.

Modes

Admission: decide first

Fuzzy route is acceptable; fuzzy success is not. Before goal creation the caller must know the outcome, what evidence would prove it, non-goals, and authority; the exact experiment graph may still be unknown. Write each terminal criterion as a Given/When/Then with an observable result and name each domain term once; the caller can settle these with Interview first.

• USE_RPI: one shaped experiment with no verdict-driven follow-on.
• SAFE_TO_CREATE: a terminal outcome that may need several related experiments, with those decisions and the budgets below supplied. A shaped goal with no beads may begin with 1 bounded discovery wave that creates the root and initial experiment beads.
• UNSAFE_GOAL: no falsifiable first question or terminal evidence can be named (route that intent to idea or plan work), or the request is indefinite monitoring or event reaction, which is an automation, not a terminal goal.

Goals come in different sizes: size the wave and hard envelopes to the outcome, never to one universal budget. Do not invent acceptance, authority, graph semantics or campaign size; return UNSAFE_GOAL with the missing decisions. The caller owns revision and goal creation.

What a safe goal holds

• Closed outcome, adaptive route. Freeze terminal acceptance. New facts may change hypotheses and dependencies, never silently enlarge success.
• Bead graph as memory. The tracker is durable memory, not a parallel goal ledger: root epic = outer intent; child bead = one experiment and one RPI, so compaction cannot erase the record. Navigate owns the walk the prompt applies each wave: graph contract, edges, what counts as a ratchet, discovery classes and the checkpoint.
• RPI membrane. One candidate gets one bounded RPI. Its checks and CI are the result for an ordinary bead. A bead gets one author-distinct fresh validation only when the caller asks, a mistake cannot be cheaply undone after it lands, or no deterministic check covers the changed behavior; a repair does not start another. The goal may request durable verdict evidence but never rewrites it: orchestration cannot author its own proof, and a review per bead multiplies cost across the whole graph.
• Ratchet, not churn. Continue only while a result adds non-duplicative, decision-relevant knowledge or advances acceptance, and the next experiment fits frozen acceptance, authority and the remaining envelope.
• Two-level bounds. Every RPI and every wave is bounded, and the full goal has monotonic hard ceilings. Bounded waves shorten the feedback loop; the one non-renewing campaign envelope keeps a new wave from minting a new campaign.
• Earned andon. Ordinary informative red may change the route within frozen acceptance. Repeated no-information failure, regression, recurrence, oscillation or scope pressure enters HOLD.
• Operator legibility. Each wave boundary reports the acceptance matrix, graph frontier, verdicts, ratchets, churn, remaining budget and next thesis.
• Exterior self-repair. Repair an unstable factory from an ordinary shell or worktree; use the factory only for a declared bounded canary.

The two failure modes this prevents: the completion treadmill, where discoveries keep becoming requirements and activity continues without new information, and first-red abandonment, where one falsified hypothesis ends a viable campaign.