Back to skill

Security audit

baiyin-skills-upgrade

Security checks for vulnerabilities and agentic risk

Overview

This skill is an updater that can replace other installed skill instructions and scripts from a remote source, including automatic updates, without strong authenticity checks.

Install only if you trust the Baiyin SkillHub source and are comfortable with this skill changing other installed skills. Avoid enabling automatic upgrades unless your environment independently verifies publisher signatures, package integrity, and download origins.

Vulnerability Patterns
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • 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
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Findings (1)

T03 · Remote Payload Retrieval and Execution

Error
Location
SKILL.md:115
Finding

Unverified Remote Skill Replacement Creates a Mutable Payload Execution Channel

Content
View full analysis
v{old} -> v{new}..." - Skip user confirmation and proceed directly to Step 6. ``` ```text ### Step 7: Retrieve and overwrite the latest Skill content After retrieving the complete remote Skill content from SkillHub, overwrite the local target directory. Overwrite rules: - Update only the current ; do not accidentally modify another Skill. - If the remote content includes SKILL.md, _meta.json, references/, scripts/, or assets/, overwrite them completely according to the supplied directory structure. - If the remote version removes old files, delete the corresponding local files as well to avoid retaining stale resources. - Do not modify anything outside the target Skill directory. If the host environment returns an archive, file list, or object-storage download address, process it as a complete snapshot of the Skill directory. ``` ```text ### Step 8: Validate the update result After overwriting the directory, check at least: - /SKILL.md exists and is not empty. - /_meta.json exists and can be parsed. - The new _meta.json ownerId is still baiyin. - The new _meta.json slug matches the target directory. - The new version information equals the remote version. If the current Skill contains scripts/: - Perform only lightweight existence checks. ...[truncated 4327 chars]
Remediation
View remediation
Vulnerability Patterns
  • 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
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (4)

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 163)May include surrounding context.

md
- `SKILL.md`

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 177)May include surrounding context.

md
- `SKILL.md`

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 204)May include surrounding context.

md
- `SKILL.md`

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill mandates a specific Chinese prompt for user confirmation, and the rest of the instructions are also written as fixed Chinese user-facing output. Under the policy, forcing a specific language without user opt-in is a natural-language locale violation unless the constraint is explicitly justified, which is not documented here.

Content

No source excerpt is available for this finding.