AgentStack
SKILL verified MIT Self-run

Log Rotation

skill-jasonkneen-lazar-log-rotation · by jasonkneen

A Claude skill from jasonkneen/lazar.

No reviews yet
0 installs
9 views
0.0% view→install

Install

$ agentstack add skill-jasonkneen-lazar-log-rotation

✓ scanned · ✓ verified — works with Claude Code, Cursor, and more.

Security review

✓ Passed

No issues found. Passed automated security review. · v0.1.0 How review works →

  • Prompt-injection patterns
  • Secret / credential exfiltration
  • Dangerous shell & filesystem operations
  • Untrusted network calls
  • Known-malicious package signatures

What it can access

  • Network access No
  • Filesystem access No
  • Shell / process execution No
  • Environment & secrets No
  • Dynamic code execution No

From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.

Are you the author of Log Rotation? Claim this listing to set pricing, connect Stripe payouts, and keep 70% of every sale.
Sign up to claim

About

log-rotation

Add a richer summary to a log archive — top user prompts, top tool commands, file-touch index — beyond what the kernel writes automatically. Optionally compress old archives and prune very old ones.

The kernel auto-rotates stream.jsonl when it exceeds LAZAR_LOG_MAX_BYTES (default 10MB). That's the safety floor — you do NOT need this skill to keep the log from breaking the agent. The kernel writes a minimal summary (archive name, size, timestamp) into memory/log-summaries/.md whenever it rotates.

This skill is for the polish layer on top: when you want a per-archive summary that includes top-30 user prompts and top-20 tool commands, or you want to compress/prune archives.

When to use

  • After kernel auto-rotation, if you want richer summaries than the minimal kernel header.
  • When you want to compress old .bak archives to save disk.
  • When you want to prune the oldest archives (>5 in keep-list).
  • The user explicitly asks "summarize the last session" or similar.

Do NOT run this on every prompt. Once-per-archive is plenty. The kernel's auto-rotation keeps the log safe regardless.

Hard rules

  • Rotation itself is non-destructive — moves the live log to .bak, never deletes.
  • mv is atomic on the same filesystem — safe under concurrent writes. The next append_stream call from the kernel creates a fresh empty stream.jsonl.
  • Don't manually rotate just for the sake of it. Let the kernel handle the size threshold; this skill enriches archives, it doesn't replace the floor.

Recipes

Check size (cheap, do this first)

LOG=$LAZAR_HOME/logs/stream.jsonl [ -f "$LOG" ] && wc -c "$LOG"

Rotate if over threshold (and write a summary to memory/)

LOG=$LAZAR_HOME/logs/stream.jsonl THRESHOLD=10485760 # 10 MB

if [ -f "$LOG" ]; then SIZE=$(stat -f%z "$LOG" 2>/dev/null || stat -c%s "$LOG" 2>/dev/null) if [ "${SIZE:-0}" -gt "$THRESHOLD" ]; then TS=$(date +%s) ARCHIVE="${LOG}.${TS}.bak" mv "$LOG" "$ARCHIVE" echo "rotated: ${LOG} -> ${ARCHIVE} (was ${SIZE} bytes)"

# CRITICAL: write a summary into memory/ so long-term recall stays cheap. # This is the L2 tier — the agent reads these instead of scanning .bak files. mkdir -p $LAZARHOME/memory/log-summaries SUM=$LAZARHOME/memory/log-summaries/${TS}.md FIRSTTS=$(head -n 1 "$ARCHIVE" | jq -r '.tsms // empty' 2>/dev/null || echo "?") LASTTS=$(tail -n 1 "$ARCHIVE" | jq -r '.tsms // empty' 2>/dev/null || echo "?") NINVOKES=$(grep -c '"kind":"invokestart"' "$ARCHIVE" 2>/dev/null || echo 0) NTOOLS=$(grep -c '"kind":"toolresult"' "$ARCHIVE" 2>/dev/null || echo 0)

cat > "$SUM" /dev/null || date +%Y-%m-%d)

archive: $ARCHIVE size: ${SIZE} bytes firsttsms: ${FIRSTTS} lasttsms: ${LASTTS} invocations: ${NINVOKES} toolcalls: ${N_TOOLS}

User prompts (last 30)

EOF jq -r 'select(.kind == "user") | .content' "$ARCHIVE" 2>/dev/null \ | tail -n 30 \ | sed 's/^/- /' \ >> "$SUM" || true

echo "" >> "$SUM" echo "## Tool commands (top 20 by frequency)" >> "$SUM" echo "" >> "$SUM" jq -r 'select(.kind == "tool_result") | .command' "$ARCHIVE" 2>/dev/null \ | awk '{print $1}' \ | sort | uniq -c | sort -rn \ | head -n 20 \ | sed 's/^/ /' \ >> "$SUM" || true

echo "summary: $SUM" else echo "no rotation needed (${SIZE:-0} bytes, threshold ${THRESHOLD})" fi fi

Compress old archives (optional, saves disk)

for f in $LAZAR_HOME/logs/stream.jsonl.*.bak; do [ -f "$f" ] && [ ! -f "${f}.gz" ] && gzip "$f" done

Prune very old archives (optional, conservative — keeps last 5)

ls -t $LAZAR_HOME/logs/stream.jsonl..bak 2>/dev/null | tail -n +6 | xargs -r rm

Pairing with load-context and archive-search

The three tiers of memory:

  • L1: $LAZAR_HOME/logs/stream.jsonl (current, raw) — bounded tail/grep via _meta/load-context.
  • L2: $LAZAR_HOME/memory/log-summaries/.md (durable summaries this skill writes at rotation) — read these first when looking back, they tell you which archive to dig into.
  • L3: $LAZAR_HOME/logs/stream.jsonl.*.bak (full forensic record) — searched via _meta/archive-search only when L1+L2 don't have what's needed.

Never cat an archive. Always grep with bounds. If you compress archives with gzip, note it in memory so future-you uses zgrep/zcat.

Principle

The kernel just records. Hygiene is the agent's responsibility, expressed as this skill. Treat rotation like rotating syslog or nginx access logs — boring infrastructure, but boringly reliable.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

Install and usage instructions live in the source repository linked above.

Reviews

No reviews yet — be the first.

Versions

  • v0.1.0 Imported from the upstream source.