Laurels

Laurels

Laurels record work that turned out genuinely good, surfaced back at session start as calibration ("this approach worked — reuse it"), not vibes. A laurel grants no task and no priority; there is nothing to farm.

Capture (during a session)

When you notice a genuinely good result — a satisfying fix, a non-obvious approach that paid off — nominate it, sparingly:

python3 ~/.claude/skills/laurels/scripts/laurels.py add "<what worked + why>" --cwd "$PWD"

Not every passing test. Only work worth remembering weeks later. Over-nomination burns Les's adjudication attention — that is the scarce resource, so err toward restraint.

Adjudicate (at session-wrapup)

session-wrapup runs this as a shared-spine step; you rarely invoke it standalone.

laurels.py pending --cwd "$PWD"      # this project's candidates, with indices
laurels.py accept <index...>         # move blessed ones into the pool
laurels.py drop <index...>           # discard the rest

Surface (at session start)

A SessionStart hook runs laurels.py show --cwd "$PWD" and injects a project-relevant laurel or two as calibration. Nothing to act on when you see them.

Store

~/.claude/laurels/pending.md (nominations) and laurels.md (accepted pool). Runtime data, project-tagged, hand-editable. Override location with LAURELS_DIR.