registry / AI-agent capability abuse / @capgo/cli / 8.48.0

@capgo/cli@8.48.0

A CLI to upload to capgo servers

AI Security Review

scanned 2mo ago · by lpm-firewall-ai

The package is not malicious by the inspected evidence, but its prompted global coding-agent extension setup is a meaningful capability that warrants a warning. Scanner claims of lifecycle abuse and remote code execution were not confirmed.

Static reason

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

Trigger

A user runs the interactive CLI flow and accepts the coding-agent skills prompt.

Impact

May modify a coding agent's global skill configuration through the external skills command.

Mechanism

Prompted global coding-agent skill installation

AI rationale

Evidence

dist/index.js

Network endpoints1

github.com/Cap-go/capgo-skills

Decision evidence

public snapshot

AI called this Suspicious at 80.0% confidence as Dangerous Capability with medium false-positive risk.

Evidence for warning

The CLI contains an interactive flow that offers to install Capgo skills globally for a coding agent and runs a package-manager skill command after confirmation.
The package exposes a Node command-line entrypoint, so this capability is reachable when users run the CLI.

Evidence against

package.json has no preinstall, install, postinstall, or prepare lifecycle hook.
The global skill setup is guarded by a TTY check and an interactive prompt.
The reported eval and Function matches are bundled dependency code; no remote payload decoding and execution was confirmed.
No credential or environment-value exfiltration was confirmed from the inspected source.

Behavioral surface

Source

ChildProcessEnvironmentVarsNetwork

Supply chain

HighEntropyStringsMinifiedUrlStrings

ManifestNo manifest risk signals triggered.

scanned 0 file(s), 512 KB of source, external domains: capgo.app, console.capgo.app, developer.android.com, eu.i.posthog.com, github.com, semver.org

Source & flagged code

17 flagged

dist/src/sdk.jsView file

102`)};qv0.renderString=function(D,Q,X){var J="",Y;if(typeof D!=="string"||D==="")return"";if(D[0]!=="{"&&D[0]!=="["){var Z;if(D.indexOf("{")===-1)Z=D.indexOf("[");else if(D.indexOf("... L103: `,D=D.substr(Z)}try{Y=JSON.parse(D)}catch(F){return u8.red("Error:")+" Not valid JSON!"}return J+=qv0.render(Y,Q,X),J}});var iB=E((Ll)=>{Object.defineProperty(Ll,"__esModule",{valu... L104: In order to be iterable, non-array objects must have a [Symbol.iterator]() method.`)}function $k0($,D){var Q=$==null?null:typeof Symbol<"u"&&$[Symbol.iterator]||$["@@iterator"];if(... ... L112: `)}),D}});var Qi=E(($f1,Di)=>{var v7=Q0("constants"),wh0=process.cwd,RH=null,Bh0=process.env.GRACEFUL_FS_PLATFORM||process.platform;process.cwd=function(){if(!RH)RH=wh0.call(proces... L113: GFS4: `),console.error($)};if(

Critical

Remote Asset Decode Execute

Source fetches a remote non-code asset, decodes its contents, and dynamically executes the decoded payload.

dist/src/sdk.jsView on unpkg · L102

112`)}),D}});var Qi=E(($f1,Di)=>{var v7=Q0("constants"),wh0=process.cwd,RH=null,Bh0=process.env.GRACEFUL_FS_PLATFORM||process.platform;process.cwd=function(){if(!RH)RH=wh0.call(proces... L113: GFS4: `),console.error($)};if(!c$[vD]){if(kR=global[vD]||[],Ui(c$,kR),c$.close=function($){function D(Q,X){return $.call(c$,Q,function(J){if(!J)Gi();if(typeof X==="function")X.appl... L114: ${Q.trim()}`:"";return`This Capgo API requires @capgo/cli@${D} or newer. You are using @capgo/cli@${$}.${X}

High

Child Process

Package source references child process execution.

dist/src/sdk.jsView on unpkg · L112

102`)};qv0.renderString=function(D,Q,X){var J="",Y;if(typeof D!=="string"||D==="")return"";if(D[0]!=="{"&&D[0]!=="["){var Z;if(D.indexOf("{")===-1)Z=D.indexOf("[");else if(D.indexOf("... L103: `,D=D.substr(Z)}try{Y=JSON.parse(D)}catch(F){return u8.red("Error:")+" Not valid JSON!"}return J+=qv0.render(Y,Q,X),J}});var iB=E((Ll)=>{Object.defineProperty(Ll,"__esModule",{valu... L104: In order to be iterable, non-array objects must have a [Symbol.iterator]() method.`)}function $k0($,D){var Q=$==null?null:typeof Symbol<"u"&&$[Symbol.iterator]||$["@@iterator"];if(... ... L111: L112: `)}),D}});var Qi=E(($f1,Di)=>{var v7=Q0("constants"),wh0=process.cwd,RH=null,Bh0=process.env.GRACEFUL_FS_PLATFORM||process.platform;process.cwd=function(){if(!RH)RH=wh0.call(proces... L113: GFS4: `),console.error(

High

Same File Env Network Execution

A single source file combines environment access, network access, and code or shell execution; review context before blocking.

dist/src/sdk.jsView on unpkg · L102

102`)};qv0.renderString=function(D,Q,X){var J="",Y;if(typeof D!=="string"||D==="")return"";if(D[0]!=="{"&&D[0]!=="["){var Z;if(D.indexOf("{")===-1)Z=D.indexOf("[");else if(D.indexOf("... L103: `,D=D.substr(Z)}try{Y=JSON.parse(D)}catch(F){return u8.red("Error:")+" Not valid JSON!"}return J+=qv0.render(Y,Q,X),J}});var iB=E((Ll)=>{Object.defineProperty(Ll,"__esModule",{valu... L104: In order to be iterable, non-array objects must have a [Symbol.iterator]() method.`)}function $k0($,D){var Q=$==null?null:typeof Symbol<"u"&&$[Symbol.iterator]||$["@@iterator"];if(... L105: In order to be iterable, non-array objects must have a [Symbol.iterator]() method.`)}var Y=!0,Z=!1,F;return{s:function(){Q=Q.call($)},n:function(){var U=Q.next();return Y=U.done,U}... L106: `)}if(!D||typeof D!=="object")D={}

High

Command Output Exfiltration

Source combines command execution, command-output handling, and outbound requests; review data flow before blocking.

dist/src/sdk.jsView on unpkg · L102

High

Obfuscated Payload Loader

Source contains an obfuscated payload loader that reconstructs and executes hidden code.

dist/src/sdk.jsView on unpkg · L112

1import{createRequire as $j0}from"node:module";var module={exports:{}};var rE0=Object.create;var{getPrototypeOf:oE0,defineProperty:lF,getOwnPropertyNames:Kb}=Object;var G2=Object.pr... L2: `)}var lq="\x1B",Tb="",zj0=39,V2="\x07",Cb="[",Lj0="]",Pb="m",O2,Eb,jb=($)=>{if($>=30&&$<=37)return 39;if($>=90&&$<=97)return 39;if($>=40&&$<=47)return 49;if($>=100&&$<=107)return... ... L7: `),X=D.split(` L8: `),J=Math.max(Q.length,X.length),Y=[];for(let Z=0;Z<J;Z++)Q[Z]!==X[Z]&&Y.push(Z);return{lines:Y,numLinesBefore:Q.length,numLinesAfter:X.length,numLines:J}}function r0($){return $==... L9: `).map((F,G,U)=>{let W=Y?Y(F,G):F;return G===0?`${X}${W}`:G===U.length-1?`${J}${W}`:`${Q}${W}`}).join(` ... L14: `).slice(Y);this.output.write(Z.join(` L15: `)),this._prevFrame=$;return}}this.output.write(iD.e

Low

Weak Crypto

Package source references weak cryptographic algorithms.

dist/src/sdk.jsView on unpkg · L1

dist/index.jsView file

5`){if(Y)$+=jTA("");let q=J?RTA(J):void 0;if(J&&q)$+=ETA(q)}else if(U===` L6: `){if(J&&RTA(J))$+=ETA(J);if(Y)$+=jTA(Y)}}return $},nVQ;var BJA=cA(()=>{t9A();QJA=`${lVQ}8;;`,NTA=new RegExp(`(?:\\${PTA}(?<code>\\d+)m|\\${QJA}(?<uri>.*)${AJA})`,"y"),nVQ=/\r?\n/}... L7: `),Y=0,Z=A;for(let G of J){if(Z<=G.length)break;Z-=G.length+1,Y++}for(Y=Math.max(0,Math.min(J.length-1,Y+B)),Z=Math.min(Z,J[Y].length)+Q;Z<0&&Y>0;)Y--,Z+=J[Y].length+1;for(;Z>J[Y].... L8: `),$=Q.split(` L9: `),J=Math.max(B.length,$.length),Y=[];for(let Z=0;Z<J;Z++)B[Z]!==$[Z]&&Y.push(Z);return{lines:Y,numLinesBefore:B.length,numLinesAfter:$.length,numLines:J}}function RB(A){return A==... L10: `).map((X,G,K)=>{let W=Y?Y(X,G):X;return G===0?`${$}${W}`:G===K.length-1?`${J}${W}`:`${B}${W}`}).join(` ... L15: `).slice(Y);this.outpu

Critical

Credential Exfiltration

Source appears to send environment or credential material to an external endpoint.

dist/index.jsView on unpkg · L5

1839contains invisible/control Unicode U+200D (zero width joiner) If you do not pass a repository name, this defaults to ${uLA} in the Cap-go org.`).action(MzQ);q6.command("star-all [repositories...]").description("⭐ Star all Capgo GitHub repositories with a small random delay between each request.\n\nIf

Critical

Trojan Source Unicode

Source contains bidi control or invisible Unicode characters associated with Trojan Source attacks.

dist/index.jsView on unpkg · L1839

5Trigger-reachable chain: manifest.main -> dist/index.js L5: `){if(Y)$+=jTA("");let q=J?RTA(J):void 0;if(J&&q)$+=ETA(q)}else if(U===` L6: `){if(J&&RTA(J))$+=ETA(J);if(Y)$+=jTA(Y)}}return $},nVQ;var BJA=cA(()=>{t9A();QJA=`${lVQ}8;;`,NTA=new RegExp(`(?:\\${PTA}(?<code>\\d+)m|\\${QJA}(?<uri>.*)${AJA})`,"y"),nVQ=/\r?\n/}... L7: `),Y=0,Z=A;for(let G of J){if(Z<=G.length)break;Z-=G.length+1,Y++}for(Y=Math.max(0,Math.min(J.length-1,Y+B)),Z=Math.min(Z,J[Y].length)+Q;Z<0&&Y>0;)Y--,Z+=J[Y].length+1;for(;Z>J[Y].... L8: `),$=Q.split(` L9: `),J=Math.max(B.length,$.length),Y=[];for(let Z=0;Z<J;Z++)B[Z]!==$[Z]&&Y.push(Z);return{lines:Y,numLinesBefore:B.length,numLinesAfter:$.length,numLines:J}}function RB(A){return A==... L10: `).map((X,G,K)=>{let W=Y?Y(X,G):X;return G===0?`${$}${W}`:G===K.length-1?`${J}$

Source & flagged code

17 flagged · loading source

dist/src/sdk.jsView file

Critical

Remote Asset Decode Execute

Source fetches a remote non-code asset, decodes its contents, and dynamically executes the decoded payload.

dist/src/sdk.jsView on unpkg · L102

High

Child Process

Package source references child process execution.

dist/src/sdk.jsView on unpkg · L112

Findings

4 Critical10 High5 Medium4 Low

CriticalCredential Exfiltrationdist/index.js

CriticalRemote Asset Decode Executedist/src/sdk.js

CriticalTrojan Source Unicodedist/index.js

CriticalTrigger Reachable Dangerous Capabilitydist/index.js

HighChild Processdist/src/sdk.js

HighSame File Env Network Executiondist/src/sdk.js

HighCredential Redirect Persistencedist/index.js

HighCommand Output Exfiltrationdist/src/sdk.js

HighObfuscated Payload Loaderdist/src/sdk.js

HighCross File Remote Execution Contextdist/index.js

HighTrigger Reachable Credential Exfiltrationdist/index.js

High

Affected versions and remediation

This report applies to @capgo/cli@8.48.0.

See all versions of @capgo/cli for other recorded verdicts.

Review the evidence and your use of @capgo/cli@8.48.0 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: 2026-09-08 11:35:55.126Z (UTC).

Think this verdict is wrong? Dispute it on GitHub or read how reports are produced.

AI Security Review

Decision evidence

Source & flagged code

Source & flagged code

Findings

Affected versions and remediation

Runtime surface

Behavioral surface

Source & flagged code