authorizelisted
Install: claude install-skill lukasrepublic/agentic-foundry
# /foundry:authorize
The operator-facing half of the front-authorization gate. The
mechanical write is performed by `${CLAUDE_PLUGIN_ROOT}/scripts/foundry-authorize.py`
(validated, logged, byte-canonical). This skill is the **operator-confirmation
wrapper** — it never re-implements the freeze logic.
## When to trigger
- Operator: "authorize `<spec>`", "/foundry:authorize `<spec-or-contract>`", "freeze the contract for `<atom>`".
- After an `acceptance-contract.yaml` is drafted (Specify-exit, a distinct authoring phase) and BEFORE any implementation dispatch.
- After a merge-time spec defect forces a re-authorization (RE-AUTH-AFTER-IMPL — pass `--reauth-after-impl`).
## Procedure
1. **Locate the pair.** Resolve the atom's spec path and its sibling
`acceptance-contract.yaml` (`specs/features/<…>/acceptance-contract.yaml`). If no
contract exists, the atom is DRAFT — it must be authored first (a distinct
contract-authoring phase; in dispatch mode by a separate `qa-engineer` worker).
2. **Review precondition (HARDENED).** Confirm the spec has cleared a review — by default,
`/foundry:spec-review` (`skills/spec-review/SKILL.md`), not the retired-as-default deep
`adversarial-spec-audit` engine (`skills/audit/SKILL.md`, kept dormant-invocable for an
exceptional deep audit only). This script's own audit-ledger precondition (`find_audit`) is
UNCHANGED and is the **NORMAL path**: it fail-closes on a spec with no matching
`.foundry/audit-ledger.jsonl` row (`s