agent-fender-skill
Claude Code skill that audits AI agent code for 7 critical safety gaps (timeout, loop limit, dangerous tools, injection, audit trail, token budget)
This report shows the evidence for agent-fender-skill@1.0.1, the only known version.
AI Security Review
scanned 4mo ago · by lpm-firewall-aiAlthough the content is product-aligned and there is no exfiltration or remote code execution, lifecycle-triggered writes into ~/.claude/skills are an unconsented mutation of a foreign AI-agent control surface. Under the install control-surface policy this is blockable AI-agent control hijack behavior.
Decision evidence
public snapshot- package.json defines postinstall: node install.js and bin points to install.js
- install.js writes package files into ~/.claude/skills/agent-fender on global postinstall
- install.js copies SKILL.md and references into Claude Code skill control surface without an interactive consent check
- SKILL.md is an agent instruction file activated by Claude Code for broad agent-code requests
- install.js skips postinstall for non-global npm installs
- No network requests, credential reads, shell execution, eval, or native/binary loading found
- Copied skill content is package-aligned safety-audit guidance rather than exfiltration or destructive instructions
Source & flagged code
2 flagged · loading sourcePackage defines install-time lifecycle scripts.
package.jsonView on unpkgInstall-time lifecycle script is not statically allowlisted and needs review.
package.jsonView on unpkgFindings
Affected versions and remediation
This report applies to agent-fender-skill@1.0.1.
The versions list below shows the verdict for each recorded version.
- Avoid installing agent-fender-skill@1.0.1. Remove it from direct dependencies and check your lockfile for transitive copies.
- Choose an independently verified alternative or release. This report does not establish that other versions are safe.
- If this version ran, investigate the affected machine and build environment. Rotate credentials it could access and rebuild from a trusted environment.
Evidence last updated: .
Think this verdict is wrong? Dispute it on GitHub or read how reports are produced.