/ce-compound

Outcome: one solved problem is written as a durable learning under <root>/solutions/, grounded against the current tree, discoverable by the next agent.

Done: the doc is written or updated, its frontmatter and claims validated, vocabulary capture recorded even when nothing qualified, and the mode's completion report emitted.

One learning per run. A session that produced several gets several sequential runs, never one batched run — references/research.md carries what batching breaks.

Setup

Run this once per invocation, before any subagent dispatch, and follow the directives it prints. Where a printed directive conflicts with this skill's own rules on asking the user questions, this skill's rules win and no blocking question is asked. Run the fence exactly as written, as its own command — no piping, filtering, truncating, or batching. Its output starts === skill context and ends with CE_CONTEXT_END. Seeing either marker without the other means the output was truncated, so run the fence again, verbatim, once. A later invocation runs its own.