Pre-Compact Resume Card

Generates an operational resume card before Claude context compaction — so the next session knows exactly where it left off, what mode it was in, and what to...

Security Scan

The skill's code and instructions are coherent with its stated purpose (generating a resume card from local transcripts and repo state); it reads and writes local project files but does not contact external endpoints or request unrelated credentials.

Purpose & Capability

The skill claims to capture pre-compaction state and the scripts implement that: they read transcripts, parse last messages, inspect git state, and write .claude/session-resume-card.md. One minor inconsistency: the registry metadata lists no required binaries, but the scripts invoke python3 and git (and bash). Those binaries are necessary for correct operation and should be declared.

Instruction Scope

The SKILL.md and scripts stay within the described purpose: backing up transcripts, extracting last user/assistant text, heuristically detecting mode, and injecting the resume card at session start. However, the scripts read a variety of project files (transcripts, .claude/*.jsonl, .claude/current-task.md, memory/MEMORY.md, brain/*.md) and list packages/apps. That is expected for a resume summary but it means the skill will access potentially sensitive project content and will write files into the repo (.claude/ and thinking/session-logs).

Install Mechanism

Instruction-only install (copy scripts or use npx clawhub installer) — no archive downloads or external installers embedded in the package. No high-risk install behavior is present in the provided files.

Credentials

The skill does not request credentials or environment variables in metadata. It does optionally honor CLAUDE_PROJECT_DIR if set. The scripts read many local files (transcripts, memory/brain notes, git state) — appropriate for the stated function but users should recognize that excerpts of recent messages and local project data will be aggregated into the resume card.

Persistence & Privilege

always:false (normal). The skill writes to its own paths (.claude/session-resume-card.md and thinking/session-logs) and does not modify other skills or system-wide settings. It can run autonomously as a hook when registered; this is expected for hook scripts.

Assessment

This appears to be what it says: a local pre-compact snapshot generator. Before installing, (1) confirm python3 and git are available in your environment (the metadata omits these required binaries), (2) inspect and, if necessary, redact sensitive content from transcripts, memory/MEMORY.md, brain/*.md, and .claude/current-task.md because the resume card will include excerpts (up to ~200 characters) and backups are kept (last 30 by default), (3) review or adjust retention/backup locations and file permissions to limit exposure, and (4) test in a non-production repo to verify behavior. There are no network calls or secret exfiltration in the provided scripts.

Claude forgets when context compacts. This skill installs a PreCompact hook that captures real operational state before compaction fires — branch, commit, last user intent, last assistant action, detected mode, and next step. The next session loads it automatically via SessionStart.

What It Does

Before every context compaction, the hook:

1. Backs up the transcript to thinking/session-logs/ (keeps last 30)
2. Parses the JSONL transcript to extract the last user intent and assistant action
3. Detects the current work mode (coding / debugging / deploy / ideation / git workflow)
4. Writes .claude/session-resume-card.md with all of this

At session start, the card is injected into context automatically — no manual copy-paste needed.

Install

npx clawhub@latest install @openclaw/pre-compact-resume-card

Or manually: copy scripts/pre-compact.sh and scripts/session-start.sh into your project's .claude/scripts/, then register the hooks in .claude/settings.json using references/settings-example.json as template.

Hook Registration

{
  "hooks": {
    "PreCompact": [{ "matcher": "", "hooks": [{ "type": "command", "command": "bash .claude/scripts/pre-compact.sh" }] }],
    "SessionStart": [{ "matcher": "", "hooks": [{ "type": "command", "command": "bash .claude/scripts/session-start.sh" }] }]
  }
}

The Insight

Written rules get forgotten. Mechanical hooks don't. This skill is the difference between Claude starting fresh (useless) and Claude starting exactly where you left off (valuable). The resume card costs ~200 tokens per session start — cheap insurance against rebuilding context from scratch.

Files

• scripts/pre-compact.sh — fires before compaction: backup + parse transcript + write resume card
• scripts/session-start.sh — inject the resume card at session start
• references/settings-example.json — hook registration template