volcengine-feedback · git:20260914.8f86bdd · 2026-09-14 · sha256 b049aa2b76ed200c
volcengine-feedback git:20260914.8f86bddA
Immutable. This exact content never changes and is served at /api/v1/blob/b049aa2b76ed200c.
--- name: volcengine-feedback description: >- Help users draft and, when authorized, submit feedback to volcengine/volcengine-skills: skill problems, missing capabilities, cloud scenarios, and successful Agent workflows. Use for explicit feedback or sharing requests, or an optional invitation at a natural summary after Volcengine CLI errors or an incompletely solved task. Feedback never blocks the main task. license: MIT --- # Volcengine Feedback Turn real experiences of managing cloud services with an Agent into actionable feedback for the developers. Welcome excitement, frustration, and hopes in the user's own words; a useful report can be a short scenario, without a diagnosis or a proposed implementation. Destination: **`volcengine/volcengine-skills`** on GitHub. [Feedback form](https://github.com/volcengine/volcengine-skills/issues/new?template=feedback.yml) · [Community and privacy guidance](https://github.com/volcengine/volcengine-skills/blob/main/.github/FEEDBACK_GUIDELINES.md). ## Keep the main task moving - An error or incomplete result is permission for an optional invitation, never permission to publish. Continue authorized diagnosis, recovery, and remaining work first. Mention feedback at a natural summary, at most once per conversation across all skills. If declined, ignored, or already offered, continue without another reminder. Routine success needs no solicitation; help share a successful scenario when the user expresses interest. - Keep an unsolicited invitation to one short sentence with the form link. Do not wait for an answer or start drafting, searching issues, collecting logs, installing tools/skills, logging in, or opening a browser just to invite feedback. Example: “这次还有未解决的部分;如果愿意,之后可以通过 [Feedback](https://github.com/volcengine/volcengine-skills/issues/new?template=feedback.yml) 分享实际场景,帮助我们改进 Agent 用云体验。” - Once the user asks for feedback, handle it as an optional side task. Missing details, unavailable GitHub access, declined publication, or submission failure never become prerequisites for cloud work. Preserve the draft in the conversation and return to remaining work. Never repeat cloud mutations or reproduce an incident just to obtain feedback evidence. ## One sentence → one preview → a result The user can simply say “帮我反馈:……” or “分享一下刚才跑通的场景”, or invoke `volcengine-feedback` in their host. This skill does not add a native `/feedback` or `ve feedback` command; use the host's actual skill invocation. Keep the interaction conversational: 1. **Draft from context.** Reuse the user's sentence and the relevant conversation. Infer the optional category (skill problem, capability request, scenario need, or success story), title, and known product/skill details. Ask at most one short question only when there is no usable experience to describe. Otherwise leave missing details out; never turn this into a questionnaire. 2. **Show one compact preview.** Present `[Feedback] <short title>` and a short paragraph explaining the goal, experience, and desired change or value, as far as they are known. Add an optional “Details” paragraph only when existing sanitized evidence helps. Versions, reproduction steps, logs, and a separate emotion rating are not required. Preserve criticism and enthusiasm in the user's language without inventing root causes, results, time savings, or maintainer commitments. 3. **Offer one publication decision.** Identify the public repository next to the exact preview and link the community/privacy guidance so this decision can cover the form's acknowledgement. The user can send it, edit it, or leave it as a draft. A drafting request does not authorize publication; use existing explicit authorization if it already covers this exact content and destination. Otherwise ask once whether to publish, without pausing unrelated main work. Example preview when the conversation supports these facts: > **[Feedback] 部署后希望 Agent 继续验证服务可用性** > > 我想让 Agent 部署一个测试应用。资源已经创建,但服务是否可用还没确认。 > 希望能继续检查并说明结果,减少在控制台之间来回排查。 > > 将公开发布到 `volcengine/volcengine-skills`,仅包含以上脱敏内容。 > 确认可公开并同意[反馈约定](https://github.com/volcengine/volcengine-skills/blob/main/.github/FEEDBACK_GUIDELINES.md)后即可发送,也可以修改或先留作草稿。 For an active incident, continue troubleshooting or official support; a public issue carries no response-time or resolution guarantee. ## Review for public sharing Before any issue search, URL prefill, attachment, or submission, minimize and redact the data. The issue will be **public**. Include only information necessary to understand the scenario: - Remove AK/SK, tokens, cookies, authorization headers, signed URLs, passwords, personal/customer data, private endpoints, local absolute paths, and confidential business content. Replace account IDs with `<account-id>`, resource IDs with prefix-preserving forms such as `vpc-<id>` and `clb-<id>`, and real TRNs with `<resource-trn>`. Check titles, screenshots and attachments too. - Prefer a short sanitized excerpt to entire transcripts, configuration files, or logs. Do not inspect credential stores or collect more cloud data for feedback. If safe redaction is uncertain, omit the detail. Security vulnerabilities or suspected credential exposure belong in the project's available private security reporting channel or official Volcengine support, not a public issue; do not invent a private contact or upload exploit details. - Follow the [GitHub Community Guidelines](https://docs.github.com/en/site-policy/github-terms/github-community-guidelines). Focus on behavior, evidence, and impact. Strong negative product feedback is welcome; personal attacks, harassment, hate, threats, and exposure of others' private information are not. Maintainers will close attacking or abusive issues directly. If a draft contains an attack, offer a factual rewrite preserving the underlying complaint; do not publish the attacking text. ## Deliver or submit Use an available authorized GitHub integration or browser after the publication decision above. There is no mandatory executable or environment-variable dependency. Form acknowledgements require the user's agreement; never tick them silently. Do not require an issue search. If a matching issue is already known or the user asks to find one, search only sanitized keywords; commenting on an existing thread requires authorization for that destination too. | Outcome | Finish the feedback side task | | --- | --- | | Submission verified by returned URL or read-back | Briefly acknowledge the contribution and return the issue link. | | Draft only, no GitHub access, or definite submission failure | Mark “not submitted”; return the copyable draft and form link without requiring login, installation, or a new conversation. Help set up access only if requested. | | Creation timed out and may have succeeded | Check for a matching issue before any retry. If unverifiable, mark “submission unconfirmed”, preserve the draft, and stop this attempt to avoid duplicates. | Continue remaining authorized main work immediately. No outcome requires waiting for a maintainer.