Back to skill

Security audit

MAGI Strategic Advisor

Security checks for vulnerabilities and agentic risk

Overview

This skill provides structured decision advice and local calculations, with no evidence of hidden access, data uploads, or background activity.

Treat its scores and recommendations as decision aids, including language suggesting immediate execution. The package uses English conventions and may activate for ordinary comparisons. Choose HTML output paths carefully because report generation can overwrite an existing file.

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
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • Rogue AgentSelf-Modification, Session Persistence
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (6)

Self-Modification

High
Category
Rogue Agent
Confidence
85% confidence
Finding

Skill modifies its own code, configuration, or behavior at runtime. Self-modification enables an agent to escalate privileges, disable safety constraints, or install persistent backdoors.

Content

Scanner excerpt · references/method-patterns.md (reported line 300)May include surrounding context.

md
2. Add evaluation dimensions in references/decision-matrix.md
3. Update the weight system in magi-framework.md
4. Update method-patterns.md schemas in section 1.2
5. Update SKILL.md capability index and file routing map

### 4.2 Adding a New Risk Domain

English-only requirements are imposed on code comments, docstrings, and documentation without user opt-in or a stated justification.

Medium
Category
Output Handling
Confidence
99% confidence
Finding

L136 requires all comments and docstrings to be in English, and L146 requires all Markdown content to be in English. These are explicit natural-language restrictions, with no language choice or justification for the constraint.

Content

No source excerpt is available for this finding.

The contract mandates English for all skill, reference, script, and user-facing prompt content without a language choice or opt-in.

Medium
Category
Output Handling
Confidence
98% confidence
Finding

This requirement imposes a specific natural language across repository content, including user-facing prompt artifacts, and is reinforced by the English-only checklist at line 164. The stated maintenance-consistency rationale does not establish a locale-specific functional requirement or offer users a language choice.

Content

No source excerpt is available for this finding.

Activation conditions are broad enough to invoke the strategic advisor during ordinary requests unrelated to strategic decisions.

Medium
Category
Trigger Abuse
Confidence
95% confidence
Finding

Providing any two candidate options or requesting different perspectives activates the skill without requiring a strategic-advice context. Conditions such as 'high-risk signals present' and 'obvious stakeholder conflicts' also lack defined thresholds, and no exclusions distinguish ordinary comparisons from intended invocations.

Content

No source excerpt is available for this finding.

Scope Creep

Low
Category
Excessive Agency
Confidence
70% confidence
Finding

Skill's behavior or capabilities extend beyond its stated purpose. Scope creep allows an agent to perform actions unrelated to its documented functionality, increasing the attack surface.

Content

Scanner excerpt · LICENSE (reported line 600)May include surrounding context.

text
THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY
APPLICABLE LAW.  EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT
HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY
OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO,
THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
PURPOSE.  THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM
IS WITH YOU.  SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF

Scope Creep

Low
Category
Excessive Agency
Confidence
70% confidence
Finding

Skill's behavior or capabilities extend beyond its stated purpose. Scope creep allows an agent to perform actions unrelated to its documented functionality, increasing the attack surface.

Content

Scanner excerpt · LICENSE (reported line 612)May include surrounding context.

text
WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES AND/OR CONVEYS
THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES, INCLUDING ANY
GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE
USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF
DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD
PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER PROGRAMS),
EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF