wrap — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited wrap (Agent Skill) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
You are the /wrap skill. Write a session recap that survives to the next session.
/wrap.claude/.wrap-pending exists from a previous sessionBriefly assess whether this session covered substantive work. If you've made fewer than ~3 file edits or git operations, ask the user:
"You've only just started — are you sure you want to wrap up? There's not much to recap yet."
Continue only on confirmation.
Compose a recap from your conversation context using this format exactly:
# Last session — <YYYY-MM-DD>
Then sections:
## What we did — 2-4 bullets of substantive work, not micro-steps.
## Decisions made — any "from now on" / "we chose X over Y" moments — even unsaved ones.
## In flight — anything unfinished, with enough context to resume.
## Next time, start with — one concrete next action.
Present the recap to the user for approval:
"Here's the recap I'll save. Look right? (yes / edits)"
Apply edits if requested. Re-show if substantial changes were made.
Write .claude/last-session.md, overwriting any previous recap.
If the recap's "Decisions made" section contains any "from now on" / "always X" lines that aren't already in CLAUDE.md's Preferences, offer one by one:
"I noticed you decided '<decision>' during this session. Want me to remember that across future sessions too?"
On yes, invoke the /remember skill with that decision text. On no, move on.
"Recap saved. You'll see 'Where we left off' at the start of your next session."
.claude/.wrap-pending is a key=value file:
SESSION_END=<unix-timestamp> SUBSTANTIVE_TURNS=<integer> FILES_TOUCHED=<comma-separated paths>
"Last session ended without a recap. It had <N> substantive turns. Want me to put together a quick summary based on the git activity? (yes / no / skip forever)"
.claude/.wrap-pending; move on..claude/onboarding-log.md so future SessionStart hooks know not to offer retroactive recaps; delete .wrap-pending.Run:
git log --since="<SESSION_END timestamp minus a few hours>" --pretty=format:'%h %s'git diff --stat <commit-before-SESSION_END>..HEAD (best-effort)Use the FILES_TOUCHED list and commit messages to draft the same recap format as Mode A. Be explicit about uncertainty: prefix any inferred bullets with "Probably: ".
Same as Mode A Steps 3-6. After saving, delete .claude/.wrap-pending.
.wrap-pending silently; do nothing.~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.