deliverlisted
Install: claude install-skill mblauberg/provenant
# Deliver
Coordinate the domain-neutral lifecycle; domain skills own their methods.
## Entry
Require an approved intent, acceptance criteria, minimum risk tier and bounded
authority. Scope/design, disclosure, one-way doors and risk downgrades need
explicit user approval. If intent is unsettled, use `scope`.
Select base profile from `config/delivery-profiles.json`: `software`,
`research`, `analysis`, `document` or `agent-product`. Add high-stakes
overlay when source authority, privacy or qualified review matters. Projects
may strengthen a profile, never weaken kernel gates silently. Use explicit
independent `not_applicable` form for fabric_relationships rather than
inventing.
## Lifecycle
1. From the authorised workspace root, create `.agent-run/<id>/RUN.json`:
```sh
"${AGENTS_HOME:-$HOME/.agents}/skills/deliver/scripts/delivery_receipt.py" init \
--run-dir ".agent-run/<id>" --run-id "<id>" --profile "<profile>" \
--chair-family "<family>" --risk-assessment "<risk-assessment.json>" \
--intent "<approved-intent-file>" --authority "<authority.json>"
```
The intent file must be non-empty; `--risk-assessment` and `--authority`
accept a JSON file path or inline JSON. Add `--fabric-relationships` only
for coordinated work, then bind other required artifacts through the same
producer and follow
[the receipt contract](references/contract.md).
2. Record each state transition. No state may jump an approval, evidence,
review, acceptance or