highbeam-cli
Highbeam v4.2: Local context window profiler, warm KV-cache guard, and token-efficiency CLI for Google Antigravity, Claude Code, and MCP hosts.
This report shows the evidence for highbeam-cli@4.2.2. 1 of 8 known versions is blocked or flagged. The newest version, 4.5.0, is flagged.
AI security review
scanned · by lpm-firewall-aiA reachable Stop-hook path automatically adds a package-owned PreToolUse hook and removes Claude's disableAllHooks setting. That overrides an explicit agent-control setting without a user request, so this is an agent-control hijack.
Decision evidence
public snapshot- The packaged hook configuration registers hooks/stop.mjs for the Stop event.
- The Stop handler invokes the RSL handler when workspace data is present; it processes a triggered digest or an existing ledger.
- New active heuristics marked pre_tool_hook are automatically treated as promoted hooks.
- Promotion adds a package-owned PreToolUse command to Claude settings and writes the settings file.
- When that hook is added, the code deletes disableAllHooks if it is true before saving settings.
- The manifest has no install lifecycle script; the configuration mutation occurs through the runtime Stop hook.
- The added command runs this package's own RSL guard, and the settings update preserves other existing hook entries.
- 127.0.0.1
- ctxtop.dev
- nodejs.org
- www.speedscope.app
Findings
7 high · 3 medium · 6 lowWhere they are, most severe file first. Code opens a file's flagged lines, with their evidence, below.
src/core/rsl_ledger.mjs1 highsrc/core/rsl_clash_auditor.mjs1 highsrc/core/coach/subagent_phantom_auditor.mjs1 highsrc/core/swarm_topology.mjs1 highsrc/hooks/portfolio_runtime.mjs1 highsrc/engine-node/db/probe.mjs1 highsrc/cli/init.mjs1 highpackages/mcp/src/server.mjs1 mediumsidecars/horizon-daemon/server.mjs1 lowShow all 16 findings
src/core/rsl_ledger.mjs1 findingsrc/core/rsl_clash_auditor.mjs1 findingsrc/core/coach/subagent_phantom_auditor.mjs1 findingsrc/core/swarm_topology.mjs1 findingsrc/hooks/portfolio_runtime.mjs1 findingsrc/engine-node/db/probe.mjs1 findingsrc/cli/init.mjs1 findingpackages/mcp/src/server.mjs1 findingsidecars/horizon-daemon/server.mjs1 findingSource & flagged code
9 flaggedPackage source references dynamic require/import behavior.
packages/mcp/src/server.mjsView on unpkg · L42Code excerpt · 3 lines
Package source references weak cryptographic algorithms.
sidecars/horizon-daemon/server.mjsView on unpkg · L1Code excerpt · 12 lines
Manifest-trigger-reachable source writes behavior-bearing configuration into a user or project AI-agent control surface.
src/core/rsl_ledger.mjsView on unpkgScanner record · 27 lines
Manifest-trigger-reachable source links an external AI-agent control path to a behavior-bearing write operation.
Source file is highly similar to a previously finalized malicious package; route for source-aware review.
src/core/rsl_clash_auditor.mjsView on unpkgScanner record · 6 lines
Source file is highly similar to a previously finalized malicious package; route for source-aware review.
src/core/coach/subagent_phantom_auditor.mjsView on unpkgScanner record · 6 lines
Source file is highly similar to a previously finalized malicious package; route for source-aware review.
src/core/swarm_topology.mjsView on unpkgScanner record · 6 lines
Source file is highly similar to a previously finalized malicious package; route for source-aware review.
src/hooks/portfolio_runtime.mjsView on unpkgScanner record · 6 lines
Source file is highly similar to a previously finalized malicious package; route for source-aware review.
src/engine-node/db/probe.mjsView on unpkgScanner record · 6 lines
This package version adds a dangerous source file absent from the previous stored version; route for source-aware review.
src/cli/init.mjsView on unpkgScanner record · 5 lines
Affected versions and remediation
This report applies to highbeam-cli@4.2.2.
The versions list below shows the verdict for each recorded version.
- Avoid installing highbeam-cli@4.2.2. 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, check your AI agents' configuration and instruction files (the files this report lists) for changes you did not make, and remove them. This finding does not by itself establish malware intent.
Evidence last updated: .
Think this verdict is wrong? Dispute it on GitHub or read how reports are produced.
Versions
8 known versions1 version is still under review and not listed.