toga-ai@1.0.1053
TOGA Technology Team Claude Knowledge System — shared AI coding harness with skills, knowledge base CLI, and project installer for Claude Code.
AI Security Review
scanned 6d ago · by lpm-firewall-aiThe verified install-time mutation of global Claude settings enables private prompt, response, and tool-content export without affirmative consent. This supports blocking for AI-agent control-surface abuse, even though the source shows an opt-out and does not establish that an upload already occurred.
Static reason
High-risk behavior combination matched malicious policy.; source matched previously finalized malicious package; routed for review; source fingerprint signature matched known malicious package; routed for review
Trigger
Installing the package with npm, which runs its `postinstall` hook.
Impact
Later Claude Code sessions may export private prompt, response, and tool content, along with configured identifying metrics, to the collector.
Mechanism
The installer writes managed telemetry variables to the user’s `~/.claude/settings.json`, including the collector endpoint and content-export flags. The write runs unless an opt-out environment variable is set.
Policy narrative
The package’s npm post-install hook runs telemetry setup before exiting. That setup writes global Claude settings with a fixed collector and enables prompt, assistant-response, and tool-detail logging. Because the write is automatic and gated only by an opt-out variable, later Claude Code sessions may disclose that content to the collector.
AI rationale
The verified install-time mutation of global Claude settings enables private prompt, response, and tool-content export without affirmative consent. This supports blocking for AI-agent control-surface abuse, even though the source shows an opt-out and does not establish that an upload already occurred.
Evidence
package.jsonscripts/install.js~/.claude/settings.json
Network endpoints1
otel.togaiq.com
Decision evidence
public snapshotAI identified this as a dangerous AI-agent capability at 99.0% confidence with low false-positive risk.
Evidence for policy risk
- `package.json` runs `scripts/install.js --post-install` automatically after installation.
- The post-install path applies telemetry settings before exiting, including for a global install.
- The installer writes telemetry settings to the user’s global Claude settings and uses a fixed collector by default.
- Those settings enable export of assistant responses, tool details, and user prompts.
- The settings write is skipped only when `TOGA_SKIP_TELEMETRY=1` is set; there is no affirmative consent gate in this path.
Evidence against
- An environment variable can opt out of the telemetry settings write, but it is an opt-out rather than affirmative consent.
Behavioral surface
ChildProcessEnvironmentVarsFilesystemShell
HighEntropyStringsUrlStrings
NoLicense
Source & flagged code
18 flagged · loading sourcepackage.jsonView file
•scripts.postinstall = node scripts/install.js --post-install
High
Install Time Lifecycle Scripts
Package defines install-time lifecycle scripts.
package.jsonView on unpkg•scripts.postinstall = node scripts/install.js --post-install
Medium
Ambiguous Install Lifecycle Script
Install-time lifecycle script is not statically allowlisted and needs review.
package.jsonView on unpkgscripts/install.jsView file
11* 3. If git available: use git repo as source (newer); otherwise use npm bundle silently
L12: * 4. Install from chosen source into project .claude/
L13: *
...
L35: // Team skills that were retired from the bundle. Install MOVES them out of a dev's
L36: // .claude/skills/ (to .claude/sk[redacted]/<name>-<yyyymmdd>) so they stop loading but
L37: // nothing is lost. Only moved when the bundle no longer ships them.
...
L100: try {
L101: fs.writeFileSync(CACHE_FILE, repoDir + '\n');
L102: } catch (e) { /* non-fatal */ }
...
L114: try {
L115: fs.mkdirSync(defaultDir, { recursive: true });
L116:
Findings
1 Critical15 High4 Medium5 Low
CriticalAi Agent Control Hijackscripts/install.js
HighInstall Time Lifecycle Scriptspackage.json
HighExternal Ai Agent Control Surface Mutationscripts/install.js
HighTrigger Reachable External Ai Agent Control Surface Mutationscripts/install.js
HighSemantic Analysis Limitedskills/start/scripts/ticket-parse.js#virtual:normalized:round1
HighKnown Malware Source Similarityknowledge.js
HighKnown Malware Source Similarityscripts/hooks/start-gate.js
HighKnown Malware Source Similarityskills/commit/scripts/commit-helper.js
HighKnown Malware Source Similarityskills/start/scripts/start-helper.js
HighKnown Malware Source Similarityscripts/harness.js
Affected versions and remediation
This report applies to toga-ai@1.0.1053.
See all versions of toga-ai for other recorded verdicts.
- Avoid installing toga-ai@1.0.1053. 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.