Security audit
Security checks for vulnerabilities and agentic risk
The skill is a document-research workflow, but it broadly auto-applies and can duplicate source files while keeping permanent process logs, so it belongs in Review before installation.
Install only if you are comfortable with the agent creating a project artifact tree, copying source documents, doing external validation, and keeping detailed process records. For sensitive, regulated, or client-confidential material, require an isolated workspace and explicit agreement on which files may be copied, what logs may contain, and when artifacts should be deleted.
The trigger language is extremely broad, including effectively any research, planning, review, consolidation, translation, or verification task, and even says to use the skill whenever the user does not mention it. That can cause unintended invocation on unrelated tasks, leading the agent to perform extensive file handling, logging, and workflow changes without the user expecting it.
The skill instructs the agent to copy original user-provided source files into an output directory and preserve originals as part of final consolidation. This duplicates potentially sensitive documents, increases data exposure, and may unintentionally include confidential inputs in shared deliverable packages.
Claims such as 'guarantees zero hallucination' and 'hallucination-free' overstate the reliability of the system and may cause users or downstream agents to trust outputs beyond what is justified. In security- or compliance-relevant workflows, misleading assurances can suppress necessary human review and verification.
The skill omits a clear upfront warning that it will create directories, copy source files, preserve artifacts, and maintain persistent logs of decisions and errors. Without disclosure, users may unknowingly expose sensitive content or authorize broader data retention than intended.
The methodology requires full decision-chain logging and traceability for every fact and deliverable, which creates a broad instruction to retain potentially sensitive user data, source contents, and internal reasoning artifacts. If applied to confidential or regulated materials, this increases the attack surface and risk of unnecessary retention or later disclosure.
Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.
FUNCTION GATE(gate\_id, user, question):
  """Mandatory checkpoint. NEVER skip or auto-approve."""
  present\_current\_state\_to\_user()
Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.
| IR-02 | Never regenerate to merge | Document fusion must use XML-level merge. Content regeneration always loses content. |
| IR-03 | Never skip verification | Every deliverable must pass Plan-vs-Content verification before being marked complete. |
| IR-04 | Never proceed past a gate without user approval | Decision gates exist to prevent wasted work. Respect them. |
The required live Process History, permanent error logs, decision logs, and file timelines instruct the agent to retain a detailed record of project actions and interaction-derived information. This can capture sensitive user inputs, operational details, and potentially confidential mistakes or decisions beyond what is necessary for task completion.
The skill defines gates as mandatory safety checkpoints that must never be skipped, but later states several gates are optional for small projects. This contradiction can cause an agent to bypass approval and verification steps inconsistently, weakening safeguards around scope, correctness, and user oversight.
No suspicious patterns detected.