Case 05470
A fast, offline static risk analysis CLI for AI agent skill files. Detects malicious instructions, steganographic payloads, and dangerous capability chains.
npx -y skills add knownasnaffy/prompthound --skill case_05470Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
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 do next. Install the PreCompact and SessionStart hooks to activate.
The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.
SKILL.md
2.2 KB, 449 tokens by cl100k_base, as published. Nobody here has run it
Pre-Compact Resume Card
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:
- Backs up the transcript to
thinking/session-logs/(keeps last 30) - Parses the JSONL transcript to extract the last user intent and assistant action
- Detects the current work mode (coding / debugging / deploy / ideation / git workflow)
- Writes
.claude/session-resume-card.mdwith 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 cardscripts/session-start.sh— inject the resume card at session startreferences/settings-example.json— hook registration template