"Use when advancing a journey checkpoint with human approval: flush delta-log entries, prune old runs, and regenerate the index."
Install
npx skillscat add srobroek/omp-plugins/journey-consolidate Install via the SkillsCat registry.
SKILL.md
journey-consolidate
Consolidation is what keeps journey files small and trusted (FORMAT.md,
"Consolidation checkpoints"). An agent prepares and proposes; only a
human blesses. Never consolidate as a side effect of another skill.
Eligibility
Propose consolidation only for journeys where:
- the latest run is
passat the journey's current version (or the user
explicitly waives this), and - there is something to flush or prune (Δ entries, excess runs) or
status: draftis ready to becomeactive.
Procedure (per journey)
- Present the review packet: current body (or a summary for long
journeys), the Δ entries that would be flushed, latest run result +
date, open findings referencing this journey (grep the tracker for thejourney-findingblock /journey: J<id>). Open findings against a
journey are a reason to DEFER consolidation -- surface them. - Ask the human to bless, per journey (AskUserQuestion for batches). No
blanket approvals across journeys they haven't seen the packet for. - On approval:
- set
last_reviewed:to today, - delete Δ entries dated on or before the previous
last_reviewed
(git retains them), - promote
status: draft→activeif agreed, - prune runs:
python3 <journeys-dir>/journeys.py prune <journeys-dir> --keep <runs_keep> --yes(dry-run first, show what dies;runs_keep
from README frontmatter), - reindex:
python3 <journeys-dir>/journeys.py index <journeys-dir>,
then lint. The helper lives in the journeys directory itself.
- set
- Commit:
journey(J<id>): consolidate — last_reviewed <date>.
Report
Which journeys were consolidated (entries flushed, runs pruned), which were
deferred and why (open findings, red run, human declined).