Back to skill

Security audit

tech-content-review-panel

Security checks for vulnerabilities and agentic risk

Overview

ClawScan rates this skill benign with medium confidence: the checks found no reported security issues, while three flagged patterns were judged expected for its stated content-review purpose; one local reference was unresolved, so the evidence is incomplete.

The verdict is benign, but not a guarantee of safety. Skillspector completed checks on four Markdown components but marked analysis partial because a local reference could not be resolved; AIG completed with no findings. Consider whether the review skill’s broad trigger suits your use before installing.

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
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • YARA SignaturesMalware Match, Webshell Match, Cryptominer Match
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (3)

YARA rule 'agent_skill_mcp_tool_poisoning_metadata': MCP/tool metadata poisoning indicators in tool schemas or skill manifests [agent_skills]

High
Category
YARA Match
Confidence
80% confidence
Finding

YARA rule matched a hack tool or exploit indicator (offensive tools, reconnaissance, privilege escalation, or exploit frameworks).

Content

Scanner excerpt · SKILL.md (reported line 3)May include surrounding context.

md
---
name: tech-content-review-panel
description: Reviews a tech/AI/industry research or in-depth analysis long-form article before publishing, via a fixed eight-role expert panel (target-reader reps, quality gatekeepers incl. fact+originality check, distribution gatekeeper) in an evaluate-then-optimize loop. Applies to tech/AI/data deep-dives, sector judgment and research pieces — not news, marketing, docs, tutorials, or short opinion posts. Trigger when the user asks to 会审/评审/review a finished deep-analysis draft or wants tech content 接近完美/可发布.
version: "1.2.5"
agent_created: true
read_when:
  - "会审

Overly Broad Trigger

Low
Category
Trigger Abuse
Confidence
70% confidence
Finding

Overly Broad Trigger: description clause 'Trigger when the user asks to 会审/评审/review a finished deep-analysis draft or wants tech content 接近完美/可发布' activates on 'to', which is too short and may match unintended inputs

Content

No source excerpt is available for this finding.

Scope Creep

Low
Category
Excessive Agency
Confidence
75% 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 · SKILL.md (reported line 165)May include surrounding context.

md
use direct statements or two complete sentences; preserve a natural contrast when it carries the judgment.
5. **Professional-vs-distribution tension is not forced into agreement** — list options for the author to decide. Principle: a hook must not sacrifice professional credibility, but must not be so professional that no one clicks.
6. **Reader-fit feedback cannot override the positioning lock** — the reading test diagnoses misunderstanding; it does not grant every role editorial authority to expand scope, change genre, add unsupported material or flatten technical depth.
7. **Do not average roles or chase unanimity** — fix factual/technical misreadings even if one role reports them; adopt comprehension fixes only when compatible with the intended reading task; park preference conflicts explicitly.
8. **Scope and length are quality constraints** — do not add examples, tables, definitions or sections merely because a reader requested them. Every addition must repair a named comprehens