This release adds 2 notable features for engineering teams evaluating rollout.
✓ No known CVEs patched in this version
Topics
+6 more
Affected surfaces
Summary
AI summarySilent ingest failures fixed by updating hook command to read JSON from stdin instead of the deprecated CLAUDE_TRANSCRIPT_PATH env var.
Full changelog
Critical fix: if you're on v0.5.1 or earlier, your ingest has silently stopped working.
Modern Claude Code changed how it passes data to SessionEnd hooks — from the CLAUDE_TRANSCRIPT_PATH environment variable to JSON on stdin. Longhand's installed hook command referenced the env var, which is now always empty, so longhand ingest-session has been failing silently on every session end. Your archive looked like it was working but it stopped growing around the time of the Claude Code change.
Run longhand doctor on v0.5.2 and it will now flag a stale hook with a red ✗ instead of hiding it.
Upgrade (re-clone / pull + reinstall)
cd /path/to/longhand
git pull
pip install -e .
longhand hook install # auto-upgrades the stale hook in settings.json
longhand doctor # verify everything is green
longhand ingest # one-shot backfill of any sessions missed during the outage
longhand hook install now detects the old env-var-based hook command and rewrites it in place — no manual JSON editing needed.
What changed
ingest-sessionreads{"transcript_path": "..."}from stdin when--transcriptis not supplied. Silently exits if neither is present, so a misconfigured hook can never crash Claude Code's hook chain.hook installauto-upgrades stale pre-0.5.2 hooks in place. Running it on an affected machine detects the old command and rewrites it to the new stdin-based form.doctorloudly flags stale hooks with a red ✗ and the upgrade instruction, instead of showing a green ✓ while ingest is broken.- 4 regression tests added to pin the stdin contract so this class of breakage is caught by CI next time Claude Code's hook API moves.
Commits
b41a969v0.5.2: fix silent hook failure on modern Claude Code7cb480aREADME: bump version to 0.5.2 and test count to 90
Stats
- 90 unit tests passing (was 86)
- Zero behavior changes for users whose hooks were still working
Weekly OSS security release digest.
The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.
No spam, unsubscribe anytime.
Share this release
About Wynelson94/longhand
Persistent local memory for Claude Code. Indexes every session JSONL verbatim into SQLite + ChromaDB for semantic recall (~126ms)
Related context
Beta — feedback welcome: [email protected]