registry / AI-agent control-surface risk / @mamadoudicko/mmt
@mamadoudicko/mmt
Compose, reuse, and run your own agent workflows (mini teams). Name them, run them on any task, watch each step.
This report shows the evidence for @mamadoudicko/mmt@0.1.1. 2 of 2 known versions are blocked or flagged. See all versions.
AI Security Review
scanned 4mo ago · by lpm-firewall-aiStatic inspection confirms unconsented npm postinstall mutation of a broad foreign AI-agent control surface, which meets the firewall block policy even without credential theft or exfiltration. The behavior is documented and package-aligned, but it occurs automatically at install time rather than via an explicit setup command. Product guard normalized a concrete AI-agent control hijack publish_block to the blockable dangerous-capability shape.
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
npm install or npm install -g of @mamadoudicko/mmt@0.1.1
Impact
Installs package-supplied Claude command instructions into ~/.claude/commands, changing future Claude Code behavior when /mmt is invoked.
Mechanism
unconsented postinstall write to AI-agent control files
Policy narrative
On installation, npm runs scripts/postinstall.js. That script copies the package's .claude/commands/mmt.md into ~/.claude/commands/mmt.md and seeds package skills/teams under ~/.my-mini-team. The installed slash command is then part of Claude Code's command surface and instructs future /mmt sessions to orchestrate subagents and write reports.
AI rationale
Static inspection confirms unconsented npm postinstall mutation of a broad foreign AI-agent control surface, which meets the firewall block policy even without credential theft or exfiltration. The behavior is documented and package-aligned, but it occurs automatically at install time rather than via an explicit setup command. Product guard normalized a concrete AI-agent control hijack publish_block to the blockable dangerous-capability shape.
Evidence
package.jsonscripts/postinstall.js.claude/commands/mmt.mdbin/mmtlib/compose.jslib/runner.jslib/store.jslib/skills.jsscripts/gen-catalog.js~/.claude/commands/mmt.md
Decision evidence
public snapshotAI identified this as a dangerous AI-agent capability at 95.0% confidence with low false-positive risk.
Evidence for policy risk
- package.json defines postinstall: node scripts/postinstall.js
- scripts/postinstall.js writes ~/.claude/commands/mmt.md during install
- scripts/postinstall.js seeds ~/.my-mini-team/skills and ~/.my-mini-team/teams at install time
- .claude/commands/mmt.md is an AI-agent slash command that instructs Claude to run subagents and write run reports
Evidence against
- scripts/postinstall.js has MMT_NO_POSTINSTALL opt-out and catches errors
- No credential harvesting or network exfiltration found in inspected source
- Network URLs are repository/homepage/catalog links, not runtime exfil endpoints
- CLI child_process usage is user-invoked for editor/claude workflows
Behavioral surface
ChildProcessEnvironmentVarsFilesystem
UrlStrings
Source & flagged code
5 flagged · loading sourcepackage.jsonView file
•scripts.postinstall = node scripts/postinstall.js
High
Install Time Lifecycle Scripts
Package defines install-time lifecycle scripts.
package.jsonView on unpkg•scripts.postinstall = node scripts/postinstall.js
Medium
Ambiguous Install Lifecycle Script
Install-time lifecycle script is not statically allowlisted and needs review.
package.jsonView on unpkgscripts/postinstall.jsView file
1Install-time AI-agent control hijack evidence:
L4: // fresh install behaves exactly like a cloned repo:
L5: // - installs the `/mmt` slash command into ~/.claude/commands/ (the runtime
L6: // that `mmt run/new/edit` drives inside an interactive Claude Code session),
...
L28: try {
L29: const src = path.join(pkg, '.claude', 'commands', 'mmt.md');
L30: if (fs.existsSync(src)) {
L31: const dstDir = path.join(home, '.claude', 'commands');
L32: const dst = path.join(dstDir, 'mmt.md');
...
L35: if (cur !== next) {
L36: fs.mkdirSync(dstDir, { recursive: true });
L37: fs.writeFileSync(dst, next);
L38: done.push(cur === null ? 'installed /mmt command' : 'updated /mmt command');
Payload evidence from .claude/commands/mmt.md:
L1: ---
L2: description: Run (or compose) a my-mini-team workflow as
Findings
1 Critical3 High3 Medium3 Low
CriticalAi Agent Control Hijackscripts/postinstall.js
HighInstall Time Lifecycle Scriptspackage.json
HighKnown Malware Source Similarityscripts/gen-catalog.js
HighKnown Malware Source Fingerprint Signaturescripts/gen-catalog.js
MediumAmbiguous Install Lifecycle Scriptpackage.json
MediumEnvironment Vars
MediumStructural Risk Force Deep Review
LowScripts Present
LowFilesystem
LowUrl Strings
Affected versions and remediation
This report applies to @mamadoudicko/mmt@0.1.1.
The versions list below shows the verdict for each recorded version.
- Avoid installing @mamadoudicko/mmt@0.1.1. 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.