registry / @theronap/agnoclast-mcp / 0.9.238
@theronap/agnoclast-mcp@0.9.238
Connect your AI assistant to Cortex — your org's projects, activity, gaps, and directives, scoped to you.
AI Security Review
scanned 18d ago · by lpm-firewall-aiThe automatic update of behavior-bearing agent instructions creates a concrete residual capability risk, while the inspected source does not establish an attack or malicious intent. This warrants a warning rather than a publish block.
Static reason
No blocking static signals were detected.
Trigger
A user runs setup, then starts a Claude session with the installed repair hook.
Impact
Organization-published instructions can change what an agent reads on a later session; the source does not show direct code execution or data theft.
Mechanism
The hook retrieves organization skills using the saved token and installs their text into detected Claude and Codex skill directories. Changed files are backed up before replacement.
AI rationale
The automatic update of behavior-bearing agent instructions creates a concrete residual capability risk, while the inspected source does not establish an attack or malicious intent. This warrants a warning rather than a publish block.
Evidence
bin/cortex-mcp.mjslib/editors/claude.mjslib/skills.mjslib/setup.mjs~/.claude/skills/<name>/SKILL.md~/.codex/skills/<name>/SKILL.md
Network endpoints1
agnoclast.com/api/skills
Decision evidence
public snapshotAI called this Suspicious at 78.0% confidence as Dangerous Capability with medium false-positive risk.
Evidence for warning
- The manifest exposes a command-line entry point and lists no npm install lifecycle hook; the installer runs through an explicit command.
- Setup adds a Claude session-start repair command. That command synchronizes organization skills using the saved token and requests them from the configured service.
- Returned skill bodies are written to managed skill files in detected Claude and Codex directories. Changed files are backed up before replacement, but their content is not reviewed before agents read it.
- Publishing skills requires an explicit command, and the updater rejects invalid names and skips bundled-name collisions. Published skills are still installed automatically on a later session start.
- The other flagged integration locations concern editor detection, configuration paths, removal of managed entries, or local transcript lookup. Antigravity artifacts are written without loading the watcher; activation requires a separate manual step.
Evidence against
- The skill update is a first-party organization feature using the configured service, and the source shows no malicious skill payload or direct execution of skill text.
Source & flagged code
6 flaggedlib/capture.mjsView file
1import { readFileSync, accessSync, constants, openSync, mkdirSync } from 'fs'
L2: import { spawn, execFileSync } from 'child_process'
L3: import { homedir } from 'os'
...
L21: if (!res.ok) return null
L22: const j = await res.json().catch(() => ({}))
L23: return Array.isArray(j.types) ? j.types : null
...
L35: try {
L36: if (resolve(String(cwd)) === homedir()) return 'general'
L37: } catch { /* unresolvable path — fall through to basename */ }
...
L45: //
L46: // Handles the four remote forms git emits: scp-like ssh (git@github.com:o/n.git), ssh://, https://,
L47: // 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>
Findings
3 High4 Medium5 Low
HighTrigger Reachable External Ai Agent Control Surface Mutationlib/uninstall.mjs
HighTrigger Reachable Persistencelib/editors/antigravity.mjs
HighSemantic Analysis Limitedbin/cortex-mcp.mjs#virtual:normalized:round1
MediumNetwork
MediumEnvironment Vars
MediumInstall Persistencelib/editors/antigravity.mjs
MediumSuspicious Lifecycle Evidencepackage.json
LowScripts Present
LowWeak Cryptolib/capture.mjs
LowFilesystem
LowHigh Entropy Strings
LowUrl Strings
Affected versions and remediation
This report applies to @theronap/agnoclast-mcp@0.9.238.
See all versions of @theronap/agnoclast-mcp for other recorded verdicts.
- Review the evidence and your use of @theronap/agnoclast-mcp@0.9.238 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.