solvelisted

# /solve — Plan-Gated Autonomous Task Execution **Conduct ALL conversation with the user in the language defined by `.claude/rules/personal-language.md`** (or the user's input language if absent). The English instructions below are for skill clarity, not for output style. Exceptions (only): tokens inside backticks or code blocks, proper nouns, ASCII acronyms. > **Tool references** (`Read`, `Edit`, `Grep`, `Glob`, `WebSearch`, the harness's question primitive, `find`, `shell`) describe **intent**, not Claude-specific tool calls. Each harness maps them to its native equivalent. The default is to solve a task end-to-end in a single ticket. **Phase 3 Plan approval is the only required user gate.** Once the Plan is approved, Phase 4 and Phase 5 run autonomously, with three narrow exceptions defined in §"Three Remaining Breakpoints" below. When delegated from `/project {slug} run` (ADR-082), `/solve` runs in **autonomous mode**: even the Phase 3 approval is replaced by a project-level execution policy + Codex PASS, and the remaining decision points queue instead of asking. See §"Autonomous Mode" below. If the session is cleared (`/clear`) and `/solve {slug}` is invoked again later, the `## Current Position` section at the top of `_task.md` plus the worktree state (for Lane B) are enough to identify the resumption point. Worst-case rework is "one step that was started but not completed". This skill is governed by `rill-autonomous-execution.md` (lane detection, Plan gate, workt