autonomous-executionlisted

# Autonomous Execution (workflow) **Primary class:** `workflow` (execution method with embedded binding predicates). **Package id** is job-named (workflow). Do not treat this package as pure policy or always-on law. See [ADR-20260801](https://github.com/SylphxAI/skills/blob/main/docs/adr/ADR-20260801-package-classes-and-standard-composition.md). **When matched:** run this workflow for the active accepted objective. ## Progressive disclosure 1. Read [references/full-standard.md](references/full-standard.md) for the full execution method, startup, graph, evidence, and completion discipline. 2. Read [references/tool-grounded-execution.md](references/tool-grounded-execution.md) when the task needs tool-grounded iteration, checkpoint, recovery, or termination semantics. 3. Read [references/bounded-objective-closure-loop.md](references/bounded-objective-closure-loop.md) when the user explicitly asks for loop engineering, a full-scope autonomous closure loop, or repeated research-execute-audit transitions for one objective. ## Soft composition (native discovery has no dependency graph) Before material mutations, open and apply only the composed packs that match the objective (do not invent a meta-router): - `decision-quality-standard` for framing, alternatives, and claim discipline - `source-authoring-standard` for Work/attempt/checkpoint/source identity - `delivery-standard` when claiming land/live done - `evidence-and-claims-standard` when grading proof str