ClawHub
Back to skill

Security audit

Decomtangle

Security checks for vulnerabilities and agentic risk

Overview

This is a documentation-only skill that teaches safer atomic tool-call sequencing and does not install tools, request permissions, or run code itself.

Installers should understand this skill changes agent behavior by encouraging more, smaller tool calls. When adapting its curl/eval examples, use only trusted local bridge endpoints, avoid exposing bearer tokens, and prefer safer native tool endpoints when available.

A.I.G

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs

SkillSpector

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (5)

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · examples/bad-mega-script-stall.md (reported line 14)May include surrounding context.

json
{
  "command": "curl -s -X POST -H \"Authorization: Bearer $TOKEN\" -d '{\"url\":\"https://www.airbnb.com/multicalendar\"}' http://<bridge-host>:3201/tools/browser/navigate && sleep 8 && for i in 1 2 3 4 5; do SNAP=$(curl -s http://<bridge-host>:3201/tools/browser/snapshot); if echo \"$SNAP\" | grep -q 'June 16, 2027'; then curl -s -X POST -d '{\"script\":\"(() => { const leaf=[...document.querySelectorAll('*')].find(e=>(e.textContent||'').includes('June 16, 2027')); leaf.closest('button').click(); })()\"}' http://<bridge-host>:3201/tools/browser/eval; break; fi; sleep 8; done && echo DONE"
}

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · examples/bad-mega-script-stall.md (reported line 14)May include surrounding context.

json
{
  "command": "curl -s -X POST -H \"Authorization: Bearer $TOKEN\" -d '{\"url\":\"https://www.airbnb.com/multicalendar\"}' http://<bridge-host>:3201/tools/browser/navigate && sleep 8 && for i in 1 2 3 4 5; do SNAP=$(curl -s http://<bridge-host>:3201/tools/browser/snapshot); if echo \"$SNAP\" | grep -q 'June 16, 2027'; then curl -s -X POST -d '{\"script\":\"(() => { const leaf=[...document.querySelectorAll('*')].find(e=>(e.textContent||'').includes('June 16, 2027')); leaf.closest('button').click(); })()\"}' http://<bridge-host>:3201/tools/browser/eval; break; fi; sleep 8; done && echo DONE"
}

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

This markdown file shows a concrete command that sends an Authorization bearer token in an HTTP request and performs browser automation over HTTP. Although the file critiques the command's reliability, it does not warn readers about credential exposure or network/data-transmission implications of reusing such a pattern.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · examples/good-multicalendar-atomic.md (reported line 79)May include surrounding context.

(A write tool takes raw content — no shell escaping exists to get wrong.)

text
exec: curl -s -X POST -H "Authorization: Bearer $TOKEN" \
      -d @/tmp/click.json http://<bridge-host>:3201/tools/browser/eval

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · references/decomposition-heuristics.md (reported line 46)May include surrounding context.

md
Count the quoting depth **within the argument value you are about to emit**.
The JSON tool-call envelope itself does not count — every call has that.
Depth 0 (no quoting inside the value) and depth 1 (quoted tokens inside a
command string, e.g. `curl -H "Authorization: Bearer $TOKEN" -d @/tmp/x.json`)
are fine. **Depth ≥ 2 — a quoted payload that itself contains quoted strings,
inside the value — is the tripwire** (e.g. a `-d '{"script":"…'…'…"}'` inline
JSON body whose script contains its own quotes). Parsers (shell, JSON, the

Static analysis

No suspicious patterns detected.