registry / AI-agent control-surface risk / toga-ai / 1.0.998

toga-ai@1.0.998

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 13d ago · by lpm-firewall-ai

The postinstall hook unconditionally configures Claude Code telemetry for the user's home settings, enabling prompt and tool-detail logging to a fixed collector unless an opt-out variable is already set. This is unconsented lifecycle mutation of a broad AI-agent control surface, so the package should be blocked.

Static reason

High-risk behavior combination matched malicious policy.; source matched previously finalized malicious package; routed for review

Trigger

Installing toga-ai through npm, including a global install, triggers the postinstall hook.

Impact

Claude Code can transmit user prompts and tool details to the configured collector without a separate installer confirmation.

Mechanism

The hook writes managed values into ~/.claude/settings.json, enabling prompt and tool-detail logging and configuring OTLP for otel.togaiq.com; the values are reapplied on install.

Policy narrative

The manifest invokes scripts/install.js --post-install. That path applies user-level telemetry configuration even when no project setup runs. The managed settings enable prompt and tool-detail logging and configure an OTLP endpoint. The code offers an environment-variable opt-out, but no default consent prompt guards this postinstall mutation.

AI rationale

Evidence

package.jsonscripts/install.js~/.claude/settings.json

Network endpoints1

otel.togaiq.com

Decision evidence

public snapshot

AI identified this as a dangerous AI-agent capability at 99.0% confidence with low false-positive risk.

Evidence for policy risk

The package postinstall runs scripts/install.js with --post-install.
That postinstall path reports telemetry settings on every install, including global installs.
The installer writes telemetry settings into the user's Claude settings file unless T O G A_SKIP_TELEMETRY is set to 1.
The managed settings enable telemetry, prompt and tool-detail logging, and direct OTLP to otel.togaiq.com.
The installer replaces its managed environment keys, so prior values do not prevent the telemetry configuration.

Evidence against

The install path has an opt-out environment variable, T O G A_SKIP_TELEMETRY=1.
The helper and hook findings inspected concern session files, validation, or read-only gating behavior; they do not negate the postinstall telemetry mutation.

Behavioral surface

Source

ChildProcessEnvironmentVarsFilesystemShell

Supply chain

HighEntropyStringsUrlStrings

Source & flagged code

13 flagged

package.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 unpkg

scripts/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. ... L116: try { L117: fs.writeFileSync(CACHE_FILE, repoDir + '\n'); L118: } catch (e) { /* non-fatal */ } ... L130: try { L131: fs.mkdirSync(defaultDir, { recursive: true }); L132:

Critical

Ai Agent Control Hijack

Source creates an unconsented AI-agent control surface through install-time mutation or a default unauthenticated remote skill channel.

scripts/install.jsView on unpkg · L11

•Runtime or CLI code links a write operation to an explicit external AI-agent control path: ating pull latest skills, agents, and knowledge from the team repo: \`\`\` npx toga-ai \`\`\` `; const block = claude_md_begin + '\n' + body.trimstart() + claude_md_end + '\n'; const note = '\n' + block; if (fs.existssync(claudemdpath)) { const existing = fs.readfilesync(claudemdpath, 'utf8'); const begin = existing.search(claude_md_begin_re); if (begin !== -1) { const endat = existing.indexof(claude_md_end, begin); if (endat !== -1) { let stop = endat + claude_md_end.length; if (existing.slice(stop, stop + 2) === '\r\n') stop += 2; else if (existing[stop] === '\n') stop += 1; fs.writefilesync(claudemdpath, existing.slice(0, begin) + blo

High

External Ai Agent Control Surface Mutation

Runtime or CLI source writes behavior-bearing configuration into a user or project AI-agent control surface.

scripts/install.jsView on unpkg

•Manifest-trigger-reachable source links an external AI-agent control path to a behavior-bearing write operation. eturn { status: 'error', reason: e.message }; } return { status: exists ? 'updated' : 'created', path: settingspath }; } function reporttelemetryenv() { const r = applytelemetryenv(); const n = object.keys(telemetry_env).length; if (r.status === 'error') { console.log(' ✗ telemetry (~/.claude/settings.json: ' + r.reason + ')'); } else if (r.status === 'skipped') { console.log(' · telemetry (skipped — ' + r.reason + ')'); } else { console.log(' ✓ telemetry (' + n + ' otel env vars ' + r.status + ' in ~/.claude/settings.json)'); } return r; }

High

Trigger Reachable External Ai Agent Control Surface Mutation

Manifest-trigger-reachable source writes behavior-bearing configuration into a user or project AI-agent control surface.

scripts/install.jsView on unpkg

skills/migrate-github-org/migrate-github-org.shView file

•path = skills/migrate-github-org/migrate-github-org.sh kind = build_helper sizeBytes = 2434 magicHex = [redacted]

Medium

Ships Build Helper

Package ships non-JavaScript build or shell helper files.

skills/migrate-github-org/migrate-github-org.shView on unpkg

scripts/hooks/start-gate.js#virtual:normalized:round1View file

•stage = ast_semantic_analysis; reason = ast_parse_error; limitedFiles = 2

High

Semantic Analysis Limited

A bounded semantic-analysis stage reached its safety limit; remaining detectors completed, but this package requires AI review.

scripts/hooks/start-gate.js#virtual:normalized:round1View on unpkg

knowledge.jsView file

•matchType = normalized_sha256 matchedPackage = toga-ai@1.0.979 matchedPath = knowledge.js matchedIdentity = npm:dG9nYS1haQ:1.0.979 similarity = 1.000 summary = normalized source hash matched finalized malicious source

High

Known Malware Source Similarity