Security audit
Security checks for vulnerabilities and agentic risk
This supplier-feedback skill exposes broader paid controls and can spend credits during actions described as previews.
Review before installing. Prefer a supplier-only version that enforces its scope and requires approval before spending. In this version, omitting --confirm does not guarantee a free preview; keep collection manual unless you explicitly want recurring charges.
scripts/ari.pyRecurring paid collection lacks an enforced confirmation gate
The account can incur recurring collection charges. The helper retrieves pricing and adds its cost notice only after submitting the schedule change.
scripts/ari.pyDeepdive's documented preview can execute paid analysis
The helper submits paid VOC analysis during an intended preview. It also overrides deepdive's deliberate quote-only downgrade when the product is absent from the main-product subscription list, potentially contradicting the emitted statement that no credits were charged.
Creating or resuming recurring ASIN watches introduces persistent financial effects outside the advertised supplier-feedback scope. The document requires explicit price acceptance, reducing the likelihood of unauthorized charges, but the mutation capability remains inconsistent with the fixed workflow.
The checklist-only description conflicts with instructions for paid automatic execution, recurring collection, and account-wide billing settings. Although the final section requires a fixed supplier workflow and explicit confirmation, competing generic instructions create an authorization ambiguity that could cause unintended credit spending; implementation-level claims cannot be verified without the referenced code.
This duplicates the recurring-watch scope issue: the skill describes creating persistent, billable monitoring despite advertising only supplier-feedback preparation. Explicit consent and accepted-price requirements mitigate the risk, but do not resolve the conflicting supplier-only and general-monitoring instructions.
The comparison workflow recommends binding a competitor and states that binding leads to weekly automatic collection. Unlike the detailed watch flow, this passage does not explicitly require a separate recurring-cost quote and authorization, creating a path from comparison analysis to unintended persistent collection.
The supplied commands concern watch configuration and pricing rather than extracting and organizing Amazon review feedback. These are undeclared capabilities with a materially different purpose. The snippet does not establish that supplier contact, purchasing, or contract actions occur, and it does not expose the underlying CLI implementation.
Persistent monitoring and recurring credit charges exceed the declared checklist-only scope. The watch workflow explicitly requires a quote and consent, which substantially reduces risk, but its inclusion conflicts with the fixed supplier-only entry point and exposes unnecessary state-changing capabilities.
The document states that new-product collection defaults to weekly, creating a persistent spending consequence beyond a one-time feedback checklist. Instructions to re-quote with manual scheduling when recurrence is not authorized are protective, but relying on the agent to override a billable default is fragile amid conflicting workflows.
The supplied command performs a category-ranking operation, not the declared extraction and organization of review issues into supplier communication materials. No subsequent checklist-generation step is shown. The implementation of ari.py is unavailable, so the snippet does not establish supplier contact, purchasing, or any particular side effect of --confirm.
The supplied commands concern watch-service pricing and watch listings rather than organizing review-derived supplier feedback. These are distinct, undeclared capabilities, not evident supporting steps for the stated feedback-list purpose. The chunk does not show supplier contact, purchasing, contract activity, or changes to watches.
The command's --type compare and --competitor arguments explicitly select competitive comparison rather than the declared review-issue feedback workflow. This supports a purpose mismatch. However, the implementation of scripts/ari.py is not supplied, so the snippet does not establish supplier contact, purchasing, contracting, liability assessment, or any particular effect of --confirm.
The supplied command explicitly selects '--type compare' and supplies a '--competitor' ASIN, indicating comparative product analysis rather than the declared review-to-supplier-feedback workflow. The script implementation is not provided, so its precise outputs and any API access cannot be verified. There is no visible evidence of supplier contact, purchasing, contracts, or liability determinations.
run_analysis replaces an absent --confirm with authorization inferred from a server-provided autoConfirm value, contradicting the explicit-confirmation contract. cmd_voc similarly permits both collection and analysis without that flag, allowing unexpected credit consumption when these commands are available; bool() also accepts nonempty strings such as 'false'.
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.
1.5.0 起服务端对前几次付费生成不看 `autoConfirmMaxCredits`，该字段**不再作为首次体验的硬拦截**，
只保留用户自设阈值与余额校验。`sampleCap` / `degraded` 表示 Free 样本封顶与轻量模型。
`voc` / `analyze` 在 autoConfirm 命中时会直接生成，返回 `autoConfirmed: true`、
`charges`（collectionCredits / analysisCredits / totalCredits / balanceBefore / balanceAfter）与 `autoConfirmNote`，
并附 `web.report` / `web.product` 网页链接。
- `voc`: Markdown VOC 报告，SSE 聚合后在 `data.content`，并归档。
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`
Without declared permissions the skill's intent is opaque and cannot be validated.
A server-provided autoConfirm flag is treated as sufficient authority to spend credits, including first-run rules that ignore the amount. User-configured thresholds can represent valid delegated consent, but server defaults alone do not establish informed spending authorization and conflict with this variant'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 is another occurrence of the same server-driven automatic-spending policy, not an independent exploit. Deferring confirmation to a response flag can bypass the supplier variant's mandatory approval step when eligibility comes from service defaults rather than the user's consent.
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`。
The workflow explicitly skips quotation and calls a command that may already collect, generate, and deduct credits before returning. Reporting charges afterward is not prior authorization, and this instruction directly conflicts with the supplier-only requirement to quote and obtain confirmation first.
1. 运行 `check`，确认账户、邮箱验证状态和可用积点。
2. 用户要 VOC / 评论分析报告时，默认运行 `voc <ASIN> --site <站点>`，**不要再先跑一次报价**。
**返回里有 `autoConfirmed: true`（或直接带 `reportId` / 正文）就说明报告已经生成并扣点了**
（1.5.0：新用户首次体验直接出结果，用户先拿到报告再谈钱）。
此时把报告讲给用户，并转述 `charges` — 本次采集扣了多少、分析扣了多少、合计多少、扣后还剩多少。
**不要在拿到结果后再补问「要不要生成」，也不要再回头补一次报价。**
The workflow directs paid keyword analysis that produces advertising terms, negative keywords, competitor-brand terms, and an Amazon backend Search Terms string. It also instructs the agent to proactively promote advertising use after a VOC report. These outputs serve advertising and listing optimization rather than the quality, packaging, and instruction feedback checklist promised at L004.
The passage directs immediate paid generation based on autoConfirm and explicitly allows initial paid runs without an amount limit. This creates unintended-credit-spending risk unless the flag reflects a previously approved delegation, which the first-run rule does not establish.
- **价格一律以服务端本次报价为准**（`analysis/quote` 与 `collect` 报价返回的 `credits` /
`totalCredits`）。文档与示例里出现的任何单价、页数、合计都只是演示，后台可调，
**不要凭记忆或旧文档报数，也不要把示例价当成承诺价。**
- 报价返回 `autoConfirm: true` 时直接生成，不要再问「要不要」。
生成后一句话交代本次实际扣了多少、扣后余额（读 `charges` 字段）。
1.5.0 起服务端对新用户前几次付费生成不看金额上限：新用户验证邮箱后只给一个 ASIN，就一次性出报告。
- 用户说「以后别问了 / 50 以内直接做」→ 运行 `autoconfirm 50`；说「以后每次先问我」→ `autoconfirm off`；