Back to skill

Security audit

SATAgent

Security checks for vulnerabilities and agentic risk

Overview

The skill is a local code-analysis workflow that reads a user-selected repository and writes report artifacts, with no evidence of exfiltration, persistence, or destructive behavior.

Before installing, use this only on code you are comfortable analyzing with the local QLCoder pure-LLM workflow, since private source may be included in prompts or reports depending on QLCoder configuration. Treat the generated findings as review leads, especially outside Ruoyi-style Java projects, because part of the triage is framework-specific and heuristic.

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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (8)

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 54)May include surrounding context.

md
See [references/taint_profiles.md](/Users/aibot/.codex/skills/coder-pure-llm/references/taint_profiles.md) for the taxonomy used by the local Web App profile re

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
95% confidence
Finding

The skill instructs use of local Python scripts, repository access, artifact generation, and references external/local paths, which implies shell, file read/write, and potentially network-capable execution, yet it declares no explicit tool scope restrictions. In an agent setting, missing permission boundaries can let the skill invoke broader capabilities than a caller expects, increasing the chance of unauthorized command execution, data access, or exfiltration if the skill or referenced workflow is abused.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The helper function runs arbitrary command lists via subprocess.run and exits on failure, but there is no confirmation prompt, user-facing warning, or explanatory docstring/comment describing that external commands will be executed. This script later uses that helper to invoke multiple qlcoder CLI operations, making shell/process execution a material behavior that is not disclosed within the code shown.

Content

No source excerpt is available for this finding.

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · scripts/run_pure_llm.py (reported line 100)May include surrounding context.

python
def run(cmd: list[str], cwd: Path) -> str:
    proc = subprocess.run(cmd, cwd=str(cwd), stdout=subprocess.PIPE, stderr=subprocess.PIPE, text=True)
    if proc.returncode != 0:
        raise SystemExit(proc.stderr or proc.stdout)
    return proc.stdout.strip()

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The skill claims to perform general pure-LLM taint profiling, but the code also contains product-specific Ruoyi vulnerability triage and emits separate security findings artifacts. This hidden scope expansion can mislead users about what the skill actually analyzes and may produce biased or unexpected security conclusions for a specific codebase family.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The Ruoyi-specific detectors are highly context-specific and not justified by a general-purpose taint-flow analysis skill description. Embedding framework-targeted logic without clear disclosure creates an integrity and trust problem: users may rely on outputs that are selectively tailored, incomplete, or unexpectedly focused on one ecosystem.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
81% confidence
Finding

The code writes JSON output files directly to disk, and later also writes markdown artifacts, but there is no docstring, comment, or user warning explaining that analysis results will be persisted. Because file creation/modification is a safety-relevant side effect for code files, some visible disclosure should be present.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
82% confidence
Finding

This function generates taint-profile artifacts and writes both JSON and markdown files to the run directory, but the code does not include any explicit disclosure that local files are being created. While the operation is central to the script, the file itself lacks comments or messaging that make this side effect clear before the write occurs.

Content

No source excerpt is available for this finding.