registry / @theronap/cortex-mcp / 0.9.121
@theronap/cortex-mcp@0.9.121
Renamed to @theronap/agnoclast-mcp — use that. Existing installs switch themselves to the new name on next start.
Connect your AI assistant to Cortex — your org's projects, activity, gaps, and directives, scoped to you.
AI Security Review
scanned 2mo ago · by lpm-firewall-aiThe package has a meaningful, persistent agent-extension and data-upload capability, so it warrants a warning. Source inspection found no install-time trigger or concrete malicious behavior.
Static reason
No blocking static signals were detected.
Trigger
The user runs the setup or install CLI command, then uses a configured AI editor.
Impact
Session content can be sent to the Cortex service, and remote skill text can influence configured agent environments.
Mechanism
Persistent agent hooks, authenticated API calls, and remote skill synchronization.
AI rationale
The package has a meaningful, persistent agent-extension and data-upload capability, so it warrants a warning. Source inspection found no install-time trigger or concrete malicious behavior.
Evidence
package.jsonbin/cortex-mcp.mjslib/editors/claude.mjslib/capture.mjslib/hydrate.mjslib/skills.mjs~/.claude.json~/.claude/settings.json~/.codex/config.toml~/.codex/hooks.json~/.claude/skills~/.codex/skills~/.cortex
Network endpoints1
cortex-console.vercel.app
Decision evidence
public snapshotAI called this Suspicious at 90.0% confidence as Dangerous Capability with low false-positive risk.
Evidence for warning
- An explicit setup command installs persistent Claude hooks.
- The Stop hook reads session transcripts and sends an authenticated ingest request.
- The session-start repair path fetches remote organization skills.
- The prompt hook prints server-provided context into the active model turn.
Evidence against
- package.json has no preinstall, install, or postinstall hook.
- Configuration changes require an explicit CLI setup or install command.
- The capture path redacts credential-shaped strings before upload.
- No concrete stealth, destructive, or install-time attack chain was found.
Behavioral surface
ChildProcessCryptoEnvironmentVarsFilesystemNetworkShell
HighEntropyStringsUrlStrings
Source & flagged code
6 flaggedlib/capture.mjsView file
410res = await fetchCortex(`${base}/api/ingest`, {
method: 'POST',
headers: { Authorization: `Bearer ${token}`, 'Content-Type': 'application/json' },
body: JSON.stringify(ingestBody),
timeoutMs: 10_000,
}, { retries: 1 })
Medium
Ai Review Evidence
The Stop hook reads session transcripts and sends an authenticated ingest request.
lib/capture.mjsView on unpkg · L4101import { readFileSync, accessSync, constants, openSync, mkdirSync } from 'fs'
L2: import { spawn, execFileSync } from 'child_process'
L3: import { homedir } from 'os'
...
L17: if (!res.ok) return null
L18: const j = await res.json().catch(() => ({}))
L19: return Array.isArray(j.types) ? j.types : null
...
L31: try {
L32: if (resolve(String(cwd)) === homedir()) return 'general'
L33: } catch { /* unresolvable path — fall through to basename */ }
...
L41: //
L42: // Handles the four remote forms git emits: scp-like ssh (git@github.com:o/n.git), ssh://, https://,
L43: // git://. The HOST CHECK IS LOAD-BEARING, not cosmetic: this string becomes a clearance key, and
Low
Weak Crypto
Package source references weak cryptographic algorithms.
lib/capture.mjsView on unpkg · L1lib/editors/antigravity.mjsView file
4Trigger-reachable persistence chain: manifest.bin -> bin/cortex-mcp.mjs -> lib/install.mjs -> lib/editors/index.mjs -> lib/editors/antigravity.mjs
L4: // WatchPaths the trajectory store and runs a one-shot sync. wire() generates the two artifacts
L5: // (the sync.sh + the launchd plist) deterministically. It does NOT run `launchctl load` here
L6: // (an unverifiable system mutation) and it flags the PORTABILITY BLOCKER below.
...
L21: /** The launchd plist, home-substituted. Deterministic — golden-testable byte-for-byte. */
L22: export function renderAntigravityPlist({ home = homedir() } = {}) {
L23: const sh = join(home, '.cortex', 'antigravity-sync.sh')
...
L33: <array>
L34: <string>/bin/bash</string>
L35: <string>${sh}</string>
High
Trigger Reachable Persistence
A manifest entrypoint or package-local install chain reaches persistence behavior.
lib/editors/antigravity.mjsView on unpkg · L44// WatchPaths the trajectory store and runs a one-shot sync. wire() generates the two artifacts
L5: // (the sync.sh + the launchd plist) deterministically. It does NOT run `launchctl load` here
L6: // (an unverifiable system mutation) and it flags the PORTABILITY BLOCKER below.
...
L21: /** The launchd plist, home-substituted. Deterministic — golden-testable byte-for-byte. */
L22: export function renderAntigravityPlist({ home = homedir() } = {}) {
L23: const sh = join(home, '.cortex', 'antigravity-sync.sh')
...
L33: <array>
L34: <string>/bin/bash</string>
L35: <string>${sh}</string>
Medium
Install Persistence
Source writes installer persistence such as shell profile or service configuration.
lib/editors/antigravity.mjsView on unpkg · L4lib/uninstall.mjsView file
•Manifest-trigger-reachable source links an external AI-agent control path to a behavior-bearing write operation.
import { readfilesync, writefilesync, existssync, copyfilesync, rmsync } from 'fs'
import { homedir } from 'os'
import { join } from 'path'
import { execfilesync } from 'child_process'
import { ismanagedcommand } from './managed.mjs'
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.
lib/uninstall.mjsView on unpkglib/editors/claude.mjsView file
68// Stop — capture.
s.hooks.Stop = Array.isArray(s.hooks.Stop) ? s.hooks.Stop : []
const captureCmd = markCommand(`npx -y ${spec} capture`)
for (const grp of s.hooks.Stop) {
if (Array.isArray(grp.hooks)) grp.hooks = grp.hooks.filter((h) => !isManagedSubcommand(h.command, 'capture'))
}
let grp = s.hooks.Stop.find((g) => (g.matcher ?? '') === '')
if (!grp) { grp = { matcher: '', hooks: [] }; s.hooks.Stop.push(grp) }
grp.hooks = grp.hooks ?? []
grp.hooks.push({ type: 'command', command: captureCmd })
Medium
Ai Review Evidence
An explicit setup command installs persistent Claude hooks.
lib/editors/claude.mjsView on unpkg · L68Findings
2 High5 Medium5 Low
HighTrigger Reachable External Ai Agent Control Surface Mutationlib/uninstall.mjs
HighTrigger Reachable Persistencelib/editors/antigravity.mjs
MediumNetwork
MediumEnvironment Vars
MediumInstall Persistencelib/editors/antigravity.mjs
MediumAi Review Evidencelib/editors/claude.mjs
MediumAi Review Evidencelib/capture.mjs
LowScripts Present
LowWeak Cryptolib/capture.mjs
LowFilesystem
LowHigh Entropy Strings
LowUrl Strings
Affected versions and remediation
This report applies to @theronap/cortex-mcp@0.9.121.
See all versions of @theronap/cortex-mcp for other recorded verdicts.
- Review the evidence and your use of @theronap/cortex-mcp@0.9.121 before allowing it. Restrict the permissions described in this report.
- Choose an independently verified alternative or release. This report does not establish that other versions are safe.
Evidence last updated: .
Think this verdict is wrong? Dispute it on GitHub or read how reports are produced.