Back to skill

Security audit

stranger-suite

Security checks for vulnerabilities and agentic risk

Overview

ClawScan’s verdict is benign with medium confidence: scanners found no AIG issues, while a dependency-install command triggered one disputed signature match and the review was partial because a referenced local path was missing.

The verdict is benign, but confidence is medium and the scan was partial: a local path-like reference was missing. Consider the flagged Git-based installation command before installing, and treat the verdict as an assessment rather than a guarantee of safety.

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
Vulnerability Patterns
  • YARA SignaturesMalware Match, Webshell Match, Cryptominer Match
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (1)

YARA rule 'agent_skill_remote_bootstrap_execution': Remote script or code download followed by execution/bootstrap installation [agent_skills]

High
Category
YARA Match
Confidence
85% confidence
Finding

YARA rule matched a known malware signature (reverse shell, backdoor, ransomware, C2 framework, or info stealer).

Content

Scanner excerpt · SKILL.md (reported line 23)May include surrounding context.

dbox, runs every README command literally, works around each failure the way a persistent new user would, and reports every stuck point — the exact step where a real person would likely quit — plus a corrected README draft. Stranger API hands an agent nothing but your API docs and one job against a local sandbox API, and reports every place the docs alone were not enough.

Install

bash
pip install git+https://github.com/RakTikoo/stranger-suite

This provides the fresh-eyes and stranger-api commands (standard library only).

Run

Test a project's README (the --expect marker is strongly recommended — it is the text a successful run must print):

bash
fresh-eyes path/to/project --out report/ --expect TEXT_THE_RUN_MUST_PRINT

Test API docs against a locally running sandbox API:

bash
stranger-api --docs docs.md --base-url http://127.0.0.1:8765 --out report/

Both print a short JSON summary and write a plain-language report.md into the --out folde