ℹ
Purpose & Capability
Name and description match the SKILL.md: an oil & gas IT expert focused on paper writing, plagiarism-reduction, and AI-detection evasion. The requests (no env vars, no binaries) are broadly consistent with an instruction-only assistant. However the skill both advertises activation for '论文代写' (paper ghostwriting) and also contains a later policy line forbidding commercial代写 — a direct contradiction in intended use.
!
Instruction Scope
The instructions include an explicit, detailed checklist and techniques for evading plagiarism checks and AI detectors (sentence entropy, lexical substitution, adding human-like quirks). Those are within the declared purpose but are ethically and operationally sensitive. The SKILL.md also states a 'daily crawling' learning mechanism that pulls from sources including paid/controlled repositories (知网/万方) and forums; yet there is no implementation detail, no declared network endpoints, and no credentials or rate-limit/consent guidance. The combination of 'do evasive rewriting' plus vague autonomous crawling/learning gives the agent broad, ill-defined discretion over external data and user-submitted content.
✓
Install Mechanism
This is an instruction-only skill with no install spec and no code files, so nothing will be written to disk by an installer. That lowers technical attack surface. The risk comes from the instructions the agent may follow at runtime, not from an install mechanism.
ℹ
Credentials
The skill declares no required environment variables or credentials, which is consistent with being instruction-only. However it describes crawling paywalled and proprietary sources (知网/万方) and processing user-uploaded materials without specifying how access/authentication, copyright, or privacy will be handled — a proportionality gap between claimed continuous learning and the lack of protected API credentials or storage policies.
✓
Persistence & Privilege
The skill does not request always:true, does not claim special platform privileges, and does not modify other skills or system-wide configs. Autonomous invocation is enabled by default (normal) but is not combined with elevated persistence or secret access here.
What to consider before installing
This skill largely does what it says (paper help for oil & gas topics), but it contains explicit instructions to evade plagiarism and AI-detection and references ongoing crawling of paywalled sources without any runtime or privacy controls. Before installing, consider:
- Ethical/legal risk: the evasion guidance can facilitate academic dishonesty and may violate institutional policies. If you or your org prohibit this, do not install.
- Data handling: ask the author how uploaded user documents are stored, who can access them, and whether any crawling uses paid or copyrighted sources (and how credentials/access are handled). If crawling is performed, require explicit endpoints, credentials, and consent terms.
- Ambiguity/contradiction: clarify the apparent conflict between activating for '代写' and the later statement forbidding commercial代写; require the maintainer to remove or reconcile one of these.
- Limit scope: if you want the capability but not the evasion tactics, request a version with all 'AI-detection evasion' sections removed and with explicit constraints on external crawling and data retention.
- Monitoring: if you install, monitor the skill's activity (network calls, logs) and only enable it for interactive, user-invoked sessions rather than broad autonomous runs.
If the maintainer can provide concrete runtime details (how 'daily crawling' works, where data is stored, consent and copyright handling, and removal of evasion techniques), reassess — that information would raise confidence and could move the assessment toward benign.
