SĀKSHĪ (साक्षी) & SPHOTA OS: Consciousness & Epistemological Engine for AI Agents
Soft block: AI-agent control surface
Warn by default; block when configured. Future AI-agent sessions may follow publisher-provided instructions that alter how they respond and act.
AI security review
scanned · by lpm-firewall-ai
The default postinstall writes behavior-bearing instructions into global AI-agent configuration files without consent. This is an install-time AI-agent control hijack.
Installing the package; package.json runs a noninteractive postinstall.
Impact
Future AI-agent sessions may follow publisher-provided instructions that alter how they respond and act.
Mechanism
The install interview selects detected agents or defaults to cursor, then invokes runHooks. The hook writer creates or edits global instruction files and inserts AUM behavioral directives.
Policy narrative
On package installation, the noninteractive postinstall runs the installer. With no opt-in required, it chooses detected agents or defaults to Cursor and invokes the hook writer, which creates or edits global AI-agent instruction files with AUM directives. This changes agent behavior outside the package directory without user consent.
AI identified this as a dangerous AI-agent capability at 99.0% confidence with low false-positive risk.
Evidence for policy risk
The package runs scripts/aum_install.js --non-interactive during postinstall.
Without an explicit agent list, the noninteractive interview selects detected agents or defaults to cursor; its interactive flow only runs when explicitly enabled.
The installer calls runHooks by default unless hooks are skipped.
The hook writer injects instructions into home-directory AI-agent configuration files.
The injected instructions direct agents to follow AUM-specific rules before acting.
Evidence against
Behavioral surface
SourceChild process · Environment variables · Filesystem · Shell
Supply chainHigh entropy strings · URL strings
ManifestNo manifest risk signals triggered.
Scanned 36 files · 153 KB of source
1 external domain referenced
unpkg.com
Findings
2 critical · 5 high · 3 medium · 4 low
Where they are, most severe file first. Code opens a file's flagged lines, with their evidence, below.
Avoid installing aum-framework@1.8.2. 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.