AI agent skill · by @yogsoth-ai

Decompose And Or Goal

Recursively decompose a goal into an AND/OR DAG whose leaves are executable or testable subgoals.

Archived 2026-10-23 · Source updated 2026-09-29 · License: Apache-2.0. Check the original repository for the latest version.

SKILL.md instructions

# decompose-and-or-goal

## Purpose

Recursively decompose a goal into an AND/OR DAG whose leaves are executable or testable subgoals.

## Input contract

```yaml
required: [top_goal, constraints]
optional: [actor_profile, obstacle_report, existing_subgoals]
constraints: [branch semantics, dependencies, and leaf criteria must be explicit]
```

## Procedure

1. Split the goal into necessary AND conditions and alternative OR paths.
2. Recurse until each leaf has an observable result or executable action.
3. Attach actors, dependencies, constraints, and unresolved feasibility to each branch.
4. Preserve shared subgoals as DAG nodes rather than duplicating them.

If the AND/OR decomposition is complete enough to test necessity and coverage, consider `validate-goal-tree` as the next tactic.

## Output contract

```yaml
produces: [goal_dag, branch_semantics, executable_leaves, dependency_edges, feasibility_questions]
delta_fields: [findings, hypothesis_updates, uncertainties, open_questions]
```

## Quality gates

- Every leaf traces to the top goal.
- AND/OR labels and dependency directions are unambiguous.
- Cycles and duplicate shared nodes are flagged.

## Failure and counterexamples

Do not split prose into labels without changing testability, and do not treat alternatives as simultaneous requirements.

## Provenance map

- `resolved: north-star-crystallization-and-or-decompose`