Back to skill

Security audit

python-packaging

Security checks for vulnerabilities and agentic risk

Overview

ClawScan rates this Python packaging skill benign with high confidence: it found no AIG security alerts, while the noted items concern publishing, cleanup commands, and a broad trigger rather than harmful behavior.

The verdict is benign, but review the package contents and destination before using publishing steps, since uploaded artifacts may become public. Take care with the recursive cleanup command, which removes the build and distribution directories.

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
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (6)

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
85% confidence
Finding

Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).

Content

Scanner excerpt · modules/uv-workflow.md (reported line 72)May include surrounding context.

uv build --no-isolation

Clear build cache

rm -rf build/ dist/

text

### Import Errors

The quick-start workflow publishes to PyPI without warning that package contents become publicly accessible.

Medium
Category
Output Handling
Confidence
94% confidence
Finding

The workflow instructs the agent to run uv publish, which uploads built package artifacts to PyPI. It does not warn that bundled source code, metadata, or accidentally included sensitive files can become publicly accessible, nor instruct the agent to review those artifacts before uploading.

Content

No source excerpt is available for this finding.

Publishing workflows upload package artifacts to PyPI without warning about public disclosure.

Medium
Category
Output Handling
Confidence
90% confidence
Finding

The release-triggered workflow executes uv publish, which defaults to uploading built distributions to public PyPI; the complete pipeline repeats this at lines 91–94. The skill does not warn that package contents become publicly downloadable, potentially exposing sensitive files accidentally included in the distribution.

Content

No source excerpt is available for this finding.

Publishing instructions omit a warning that package artifacts will be uploaded to public package registries.

Medium
Category
Output Handling
Confidence
94% confidence
Finding

The workflow instructs the agent to publish to TestPyPI and PyPI without warning that distribution contents become publicly accessible. It does not direct the user to review artifacts for private files or confirm the intended registry before uploading.

Content

No source excerpt is available for this finding.

Troubleshooting instructs recursive deletion of build and distribution directories without an accurate deletion warning.

Medium
Category
Output Handling
Confidence
92% confidence
Finding

The workflow runs rm -rf build/ dist/, permanently deleting both directories and their contents relative to the current working directory. The comment describes this as clearing a build cache, without disclosing that it also deletes built distribution artifacts.

Content

No source excerpt is available for this finding.

Overly Broad Trigger

Low
Category
Trigger Abuse
Confidence
70% confidence
Finding

Overly Broad Trigger: 'uv' is too short and may match unintended inputs

Content

No source excerpt is available for this finding.