Security audit
Security checks for vulnerabilities and agentic risk
This image-review skill also exposes recurring paid monitoring and can spend account credits without fresh approval.
Prefer a corrected package that enforces image-only commands and approval before spending. Omitting --confirm does not reliably prevent charges. If proceeding, set billing to ask every time and separately approve any recurring collection, which can continue consuming credits after the conversation.
scripts/ari.pyRecurring collection can be enabled without an executable confirmation gate
Recurring collection and subsequent credit consumption can continue beyond the current task without the separate confirmation promised by the Skill.
scripts/ari.pyDeepdive's quote-only default is overridden by shared automatic confirmation
A deepdive invocation documented as non-charging by default can generate and charge for a report. In the missing-product fallback, the output can additionally claim that analysis was reduced to quotation and no credits were charged.
The image-only skill includes conflicting general-purpose instructions for paid analysis, recurring collection, and caller-selected operations workflows. The final fixed-entry restriction mitigates this, but earlier claims that this package is unrestricted and should sometimes execute without confirmation create an actionable scope and authorization ambiguity.
Persistent watches introduce ongoing financial and account-state effects beyond a bounded image audit. Explicit pricing and consent requirements substantially reduce the risk, but the watch instructions still conflict with the final image-only execution boundary.
This overlaps the persistent-watch scope issue: an image-review skill instructs the agent how to create and manage recurring billable services. The documented consent checks reduce the likelihood of unauthorized charges, but contradictory scope instructions leave an avoidable execution path outside the advertised boundary.
The supplied invocation requests a category-level ranking and does not identify a main ASIN, authorized competitor pages, or an image-audit task. Although review feedback is relevant to the declared purpose, generating a Kitchen leaderboard by negative-feedback rate is a distinct, undeclared capability. The script implementation is not supplied, so additional effects cannot be established.
The supplied commands concern watch configuration and service costs, not comparison of product images, visible information, or review feedback. These capabilities are not disclosed in the declared purpose, and the chunk provides no connection establishing them as supporting implementation details of image auditing. The snippet does not establish that monitoring is activated or that any explicitly excluded advertising, inventory, ordering, or publishing activity occurs.
The competitor-binding instructions state that adding a competitor starts weekly collection, but do not require separate recurring-charge consent at that step. Authorization for a one-time comparison does not establish authorization for ongoing competitor collection, despite the broader scheduling safeguards elsewhere.
Weekly collection creates continuing financial obligations beyond the image comparison requested by this skill. Instructions to use manual collection without recurring authorization are protective, but the package still mixes recurring scheduling with an image-only boundary.
Watch creation and resumption can establish recurring charges unrelated to the declared image audit. The skill requires explicit price acceptance, so this is a conflicting-scope exposure rather than demonstrated automatic enrollment.
The supplied commands concern monitoring-service quotes and watch listings rather than image information review or photography guidance. These are undeclared capabilities, and the chunk provides no connection establishing them as supporting steps for the stated image-review workflow. The visible code does not establish that monitoring is enabled or modified.
The chunk exposes watch-monitoring and historical reporting operations rather than explicitly comparing product images or producing photography guidance. Nothing supplied connects these watch operations to image-information review, so they appear to introduce an undeclared capability. However, the underlying script and returned data are not shown; the evidence does not establish sales forecasting, inventory access, order handling, advertising, or automatic publishing.
The supplied entry point explicitly selects an alerts operation, which does not match the stated image-information review workflow. The implementation of scripts/ari.py is not provided, so the specific alert behavior, resources accessed, and side effects cannot be established.
The skill explicitly treats an unanswered earlier quote as a reason to generate a report in a later conversation, potentially charging through automatic confirmation. Silence is not authorization, and an idle product record does not establish current intent or acceptance of the current price.
The monitoring section exposes persistent account mutations and recurring charges outside the bounded image-review purpose. Its explicit quote, ownership, deletion-target, and consent safeguards reduce severity, but do not resolve the contradictory specialized execution scope.
run_analysis changes confirm to True based on a server-provided autoConfirm value, allowing chargeable analysis without the explicit flag promised by the module. cmd_voc has an equivalent bypass that can also submit paid collection; deepdive inherits run_analysis behavior, although actual CLI reachability depends on the unseen command allowlist.
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.
- 重新发布：各市场首批上架的 1.5.0 包早于免确认修复，新注册的免费版用户仍会被要求先确认报价。
1.5.1 与当前官网下发的修复版一致，市场侧以新版本号覆盖旧包。
- 免确认只认服务端报价返回的 `autoConfirm`；采集报价扣除首次采集免费页；
非美国站首次采集可用赠送积点时不再提示「赠送积点不可用」（行为与修复后的 1.5.0 相同）。
- 命令、字段与计费规则不变，不提高最低支持版本。
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.
- 行为变更：新用户（注册并验证邮箱后）第一次要 VOC，只给 ASIN 即可（站点默认美国站），
`voc` 直接执行——自动补采集（默认 3 页）、生成报告、按服务端定价扣费并归档，
不再停在报价等确认。费用在报告返回的 `charges` 字段透明展示（采集/分析/合计/余额前后）。
- 免确认与否**只认服务端报价返回的 `autoConfirm` 布尔**：服务端策略前 N 次付费生成不看金额上限
（新用户赠送积点按设计覆盖「3 页采集 + 一份 VOC」）；用户自设阈值优先、显式要求「每次先问」
时仍先报价；余额不足不自动。客户端不再从 `autoConfirmNote` 文案或 `autoConfirmRemaining`
自行推断——文案匹配会把 Free 小额提示里的「先确认」误判成用户要求每次询问，恰好挡住新用户。
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.
不再停在报价等确认。费用在报告返回的 `charges` 字段透明展示（采集/分析/合计/余额前后）。
- 免确认与否**只认服务端报价返回的 `autoConfirm` 布尔**：服务端策略前 N 次付费生成不看金额上限
（新用户赠送积点按设计覆盖「3 页采集 + 一份 VOC」）；用户自设阈值优先、显式要求「每次先问」
时仍先报价；余额不足不自动。客户端不再从 `autoConfirmNote` 文案或 `autoConfirmRemaining`
自行推断——文案匹配会把 Free 小额提示里的「先确认」误判成用户要求每次询问，恰好挡住新用户。
- 采集报价同口径扣除服务端 `trialFreePages`（首次采集试用免费页），避免新用户被误报「积点不足」。
- 非美国站：服务端 `credits.gift` 策略下用户首次采集不限站点，`credits/balance` 返回 `giftAllSites`，
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.
不再停在报价等确认。费用在报告返回的 `charges` 字段透明展示（采集/分析/合计/余额前后）。
- 免确认与否**只认服务端报价返回的 `autoConfirm` 布尔**：服务端策略前 N 次付费生成不看金额上限
（新用户赠送积点按设计覆盖「3 页采集 + 一份 VOC」）；用户自设阈值优先、显式要求「每次先问」
时仍先报价；余额不足不自动。客户端不再从 `autoConfirmNote` 文案或 `autoConfirmRemaining`
自行推断——文案匹配会把 Free 小额提示里的「先确认」误判成用户要求每次询问，恰好挡住新用户。
- 采集报价同口径扣除服务端 `trialFreePages`（首次采集试用免费页），避免新用户被误报「积点不足」。
- 非美国站：服务端 `credits.gift` 策略下用户首次采集不限站点，`credits/balance` 返回 `giftAllSites`，
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.
为 true 时赠送积点计入 `nonUsCollectCredits`、`siteNote` 为空；UK/DE/JP 卖家的第一份 VOC 同样一步直出。
- 余额不足仍在执行前拦截（402）并引导充值；定价以服务端配置为准，代码常量仅为兜底。
- 一致性：`analyze` 各类型（insight / keywords / trend / variant / compare）走同一口径，
`autoConfirm=true` 即直接生成并返回 `charges`。
- 文档：移除所有写死的单价与合计示例。采集/分析单价与页数由后台 `pricing` 配置决定，
一律以本次报价返回的 `credits` / `totalCredits` 为准，不得凭文档或记忆报数。
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.
- 能力层收窄：44 个专属包不再只是「声明窄」，包体 `scripts/ari.py` 同时按各自契约裁剪。
构建期用 AST 删除文档未承诺的命令、它们的解析器注册块，以及只被它们引用的辅助函数与
常量；专属包因此不再包含 `export`（任意路径写文件）、`autoconfirm`（撤销逐次确认）、
`workbench`（改写差评处理状态）、`alerts`（标记预警已读）、`competitors`、`schedule`、
`voc` / `analyze` / `charts` / `deepdive` 等母版命令。
- 新增运行期兜底闸：专属包在包内写入 `allowedCommands` 白名单，`ari.py`
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.
p = sub.add_parser("check", parents=[common], help="验证 Key、账户和积点余额")
p.set_defaults(fn=cmd_check)
p = sub.add_parser("autoconfirm", parents=[common],
help="免确认阈值：autoconfirm 50 / autoconfirm off / autoconfirm default；不带参数查看")
p.add_argument("value", nargs="?", default="")
p.set_defaults(fn=cmd_autoconfirm)
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.
p = sub.add_parser("autoconfirm", parents=[common],
help="免确认阈值：autoconfirm 50 / autoconfirm off / autoconfirm default；不带参数查看")
p.add_argument("value", nargs="?", default="")
p.set_defaults(fn=cmd_autoconfirm)
p = sub.add_parser("version", parents=[common], help="显示 Skill 版本与渠道（无需 Key）")
p.set_defaults(fn=cmd_version)
Without declared permissions the skill's intent is opaque and cannot be validated.
Automatic execution can be legitimate under a user-established spending limit, but this rule also accepts first-use and remaining promotional confirmations without an amount limit. A server eligibility flag is not evidence of informed spending authorization and conflicts with this package's explicit-confirmation requirement.
`voc <ASIN>`（没说站点就用默认的美国站，不要为此追问）。
命中服务端免确认规则（新用户首次体验不看金额 / 账户剩余免确认次数 / 用户自设阈值）且余额充足时，
**直接完成采集 + 生成 + 扣点并归档**，不再停在报价环节。绝不能描述成“先帮你问个价 / 只报价不扣点”。
免确认与否以返回里的 `autoConfirm` 为准，不要自己从提示文案里猜。
- 直接生成成功后，返回结果带 `charges`（采集扣点、分析扣点、合计、扣前/扣后余额）与 `autoConfirmNote`。
**先讲报告，再用一句话转述实际扣点**，不要在拿到结果后再补问「要不要生成」，也不要回头补一次报价。
- 只有在返回 `confirmationRequired: true` 时（余额不足、或用户自行设置了“每次先问”）
This instruction delegates the decision to charge to autoConfirm, including service-defined first-use exceptions rather than solely user-approved limits. That allows eligibility for automatic processing to substitute for the explicit spending consent required by the fixed image workflow.
命中服务端免确认规则（新用户首次体验不看金额 / 账户剩余免确认次数 / 用户自设阈值）且余额充足时，
**直接完成采集 + 生成 + 扣点并归档**，不再停在报价环节。绝不能描述成“先帮你问个价 / 只报价不扣点”。
免确认与否以返回里的 `autoConfirm` 为准，不要自己从提示文案里猜。
- 直接生成成功后，返回结果带 `charges`（采集扣点、分析扣点、合计、扣前/扣后余额）与 `autoConfirmNote`。
**先讲报告，再用一句话转述实际扣点**，不要在拿到结果后再补问「要不要生成」，也不要回头补一次报价。
- 只有在返回 `confirmationRequired: true` 时（余额不足、或用户自行设置了“每次先问”）
才报出 `totalCredits` 与余额并等待同意，同意后追加 `--confirm`。