Back to skill

Security audit

decision-journal

Security checks for vulnerabilities and agentic risk

Overview

The skill is rated suspicious because scanners report that it may publish session context and local metadata to GitHub issues without privacy safeguards, along with lifecycle and review features not clearly disclosed in its stated purpose.

Before installing, consider whether you are comfortable with deferred capture potentially publishing session details and local metadata to GitHub issues. The scanner recommends caution; its review was partial because several referenced paths were missing, so the evidence is incomplete.

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
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (4)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
90% confidence
Finding

The code implements content lifecycle classification rather than decision or lesson logging, deferred capture, or storage templates. These maturity stages and promotion rules are not described in the declared purpose. The function only reads the file and returns a recommended stage; it does not modify content or perform the promotion.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
88% confidence
Finding

The supplied chunk focuses on content lifecycle management rather than logging lessons or implementing the declared capture contract. Promotion between maturity stages and generation of review schedules are not described capabilities. The functions are undefined placeholders, so the chunk does not establish that any persistent changes or external scheduling actually occur.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
87% confidence
Finding

The code implements a content-review discovery capability rather than lesson logging, deferred capture, or storage templating. It scans all Markdown files under the current directory without restricting access to journal entries. Review-list generation is not described in the declared purpose. The supplied code shows no file modification or network activity.

Content

No source excerpt is available for this finding.

The capture contract publishes session context and local metadata to GitHub issues without a privacy warning.

Medium
Category
Output Handling
Confidence
90% confidence
Finding

The contract places context verbatim into GitHub issues alongside session identifiers, branch names, and optional local artifact paths, without requiring sensitive-data filtering or destination-visibility checks. Although issue creation is intended functionality, this can disclose confidential session content or identifying local metadata to unintended readers, particularly in public repositories or automated safety-net captures.

Content

No source excerpt is available for this finding.