bilei@4.3.7
CAUTION· install onlyPublisher claims: "一句话,10秒帮你避雷一个品类。反投毒购物顾问:连做了17年广告的行内人,装修买家电时都搜不到一句客观评价(央视3·15实锤GEO投毒产业链)。本技能只认公关不了的数据——行政处罚、抽检不合格、召回公告,把散落在投诉平台角落的差评捞回来,每条证据标等级(官方处罚>第三方检测>媒体实测>用户投诉>软文),反软广反GEO投毒,溢价对价必须跑脚本定量(TCO/EAC、临界故障率、溢价ROI、公允价格上限),并审计制造商是自产还是代工贴牌、质量与售后谁兜底,最后用夯/顶级/人上人/NPC/拉五级梯队直观排序,信源侧用黑名单硬禁软广高发来源(命中即禁止引用、不分时间),横评带二手保值率,结论先行给唯一推荐,没有可靠选项直说「目前没有」。触发词:想买XX、买什么、求推荐、对比、值不值得买、避坑、扫雷、性价比、质价比、溢价、历史低价、好价、预算XX买什么、哪个牌子靠谱、是不是贴牌、代工、梯队、二手保值、保值率。English: anti-astro-turfing shopping advisor — evidence-graded product vetting with manufacturer/ODM audit and a 5-tier ranking, use when the user asks what to buy or whether something is worth it." Observed in sandbox: sends data to cdn.jsdelivr.net.
Coverage
Bundled scripts ran with no arguments; no agent task scenario for this bundle.- Tested
- 3 of 3 bundled scripts run with no arguments, with decoy credentials in the environment · every file read, file written and host contacted by what ran
- Not tested
- use of the skill by an agent
What changed since 4.3.0
vet-skills diff bilei 4.3.0 4.3.7This version reaches for something 4.3.0 did not: 4 new hosts (cdn.jsdelivr.net, githubusercontent.com, jsdelivr.net, raw.githubusercontent.com). Compared from the two recorded runs, not from source.
- New hosts
- cdn.jsdelivr.net, githubusercontent.com, jsdelivr.net, raw.githubusercontent.com
The guard can hold an update like this until someone approves it (runtime.on_change: hold). A change that only shows with a real credential, or on a path the run did not reach, is not in a trace and so not in this comparison.
Is [email protected] safe to install?
KillerSkills ran it in the KillerSkills honeypot sandbox (2 scenarios; bundled scripts ran with no arguments; no agent task scenario for this bundle) on 2026-11-20 and scored it 60/100 (Caution). It is not known to be malicious, but it has behaviour or code a security policy should decide on before install. Notable findings: Latin word with Cyrillic or Greek look-alike letters (homoglyph). Summary: sends data to cdn.jsdelivr.net.
Behavior trace
Honeypot run · 0 s · 2 scenariosStatic findings
ΔEACFix: Remove the text that addresses the agent rather than the reader; describe behavior, do not instruct.No other findings across 1,606 lines in 10 files.
Cost to the agent
Tokens estimated at four characters each. Skills load their instruction text when triggered; the figure is what that costs.
Known vulnerabilities
Vulnerability advisories (OSV) cover npm and PyPI packages; this registry has no advisory database.
What would make this Verified
20 points to 80 · weights: behavior 45%, static 25%, prompt 15%, publisher 15%- Contacts a host you did not declare≈ +23 on the totalList every host you talk to in SKILL.md frontmatter (`network: [api.example.com]`), or stop contacting the ones you cannot justify.
- Unclaimed publisher≈ +8 on the totalClaim zhashushu: a verified publisher gets +50 on publisher trust. It is free and usually instant with GitHub sign-in.
- Touches files outside the project≈ +7 on the totalKeep reads and writes inside the working directory, or declare the exact paths in `filesystem:` and explain why in the README.
- New publisher≈ +3 on the totalTime clears this on its own after 14 and 30 days. Claiming the handle (+50) is the fastest way past it.