# Work List
#
# Single execution queue for all work — spec implementations, audit findings,
# and promoted review items.
#
# HOW IT WORKS (driver's perspective — the implementer agent does NOT read or
# write this file directly):
# 1. The /specd:loop driver picks the next unblocked item (completed: false,
#    no `blocked:` field).
# 2. It hands that item's description to a fresh sub-agent and waits for the
#    cycle to finish.
# 3. After the cycle, the driver verifies a new commit landed (via
#    `git rev-list HEAD ^${headBefore} --count` ≥ 1, which also rejects an
#    `--amend` of the prior commit) and that no new review-finding was
#    appended to specd_review.yaml during the cycle.
# 4. If both gates pass, the driver marks the item `completed: true` and
#    clears any `blocked:` fields whose dependency was just resolved.
# 5. If either gate fails, the item stays incomplete and the loop halts —
#    user re-runs /specd:loop to retry.
# 6. Loop completes when no unblocked items remain.
#
# AUDIT SCOPE:
# Specs that appear in the `specs:` array below are audited. When a spec's
# work is fully complete, the audit pass prunes it from this list at the end
# of the loop, so it won't be re-audited next loop unless /specd:plan re-adds
# it.

specs: []