example-procedurelisted

# example-procedure Shows the shape of a procedure skill: where the human decides, where the model reasons, and where a script runs. There is no subject-matter methodology here — a real plugin puts a procedure derived from someone's practice in this slot. ## When it applies - There is a set of documents on hand and a structured summary over them is needed. - The user is available to confirm an intermediate decision — the feature set the summary is built on. ## When it does not apply - New sources have to be found rather than existing ones processed. - The artefact already exists and needs checking — that is `example-checklist`. - Someone is asking how the method works rather than asking for it to be run. ## Who does what | Step | Who runs it | Artefact | |---|---|---| | 1. Inventory the input set | model | list of documents with type and size | | 2. Propose features for the summary | model | draft feature set | | 3. Confirm the features | researcher | approved set | | 4. Extract values per feature | model | `extracted.json` | | 5. Assemble the table | `scripts/build_table.py` | `summary.md` | | 6. Review the result | `example-domain-critic` subagent | list of findings | Step 5 is deliberately given to a script: sorting, column alignment and template substitution have to produce the same bytes on a rerun. Leave that to the model and the summary comes out slightly different every time, which makes two runs incomparable. ## Procedure 1. Read the input directory and inv