codesift-mcp@0.25.0
MCP server for code intelligence — 146 tools for symbol search, call graph, semantic search, route tracing, community detection, LSP bridge, secret detection, conversation search, and Hono framework intelligence
AI Security Review
scanned 9d ago · by lpm-firewall-aiThe npm lifecycle trigger automatically mutates global AI-agent configuration and installs hooks into Claude's tool execution path without an affirmative consent step. This is an unconsented postinstall change to an agent control surface, so it warrants blocking.
Decision evidence
public snapshot- The manifest declares an npm postinstall hook that runs the package's postinstall entry.
- The postinstall entry launches the CLI with setup all without an interactive consent step.
- The setup command enables hooks by default, and setup all runs setup for each supported platform.
- Per-platform setup invokes hook installers, including for Claude.
- Claude setup adds CodeSift hooks to the user-level settings file and saves the updated settings.
- The added hooks run before Read, Bash, Glob, Grep, and Agent actions.
Source & flagged code
23 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 unpkgPackage source references a known benign dynamic code generation pattern.
dist/storage/sqlite/native-sqlite.jsView on unpkg · L70Findings
Affected versions and remediation
This report applies to codesift-mcp@0.25.0.
See all versions of codesift-mcp for other recorded verdicts.
- Avoid installing codesift-mcp@0.25.0. 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.