run-stagelisted

# Run Stage Skill Executes a single pipeline stage by following its CONTEXT.md contract. Loads only declared inputs, produces only declared outputs, and pauses for human review before advancing. ## Trigger Use this skill when: - The user asks to run, execute, or process a specific stage (e.g., "run stage 2", "execute 01_research") - The user says "next stage", "run the next one", "process this stage" - The user wants to advance through a pipeline one stage at a time ## Procedure ### Step 1: Identify Target Stage Determine which stage to run: - If the user specifies a stage number or name, use that - If the user says "next stage", find the last stage with output files and advance to the next one - If ambiguous, list all available stages and ask ```bash # Find numbered stage folders ls -d [0-9][0-9]_*/ 2>/dev/null || ls -d stages/[0-9][0-9]_*/ 2>/dev/null ``` If no numbered stage folders exist, halt: "No pipeline stages found. Use `/pipeline-scaffold` to create one, or ensure stage folders follow the `{NN}_{name}/` naming pattern." ### Step 2: Read the Stage Contract Open `{stage-folder}/CONTEXT.md`. Parse three required sections: - **Inputs table** — files to load, their source paths, and Layer designation (3 or 4) - **Process section** — the numbered instructions to follow - **Outputs section** — what to produce and where to write it If any section is missing, halt and report the gap: - Missing Inputs: "Stage {NN} has no Inputs table. Add one to its CONTEXT.md befo