Workflow & Core Tools · dontbesilent2025/dbskill

dbs-goal

Name: dbs-goal
Author: dontbesilent2025/dbskill

Extracts existing constraints and asks follow-up questions only for information affecting execution or acceptance, organizing vague desires and goals into actionable, verifiable deliverables. Used when users request goal clarification, task clarity checks, or deliverable definitions.

L1 Basic · Source registration onlyV1 Commit-PinnedDual-pass Audit

Author dontbesilent2025GitHub dontbesilent2025/dbskill ↗Stars 10121License CC BY-NC 4.0（Attribution-NonCommercial 4.0 International；仓库根 LICENSE 明文 NonCommercial — 非商业授权，不得商用；GitHub API license=NOASSERTION/Other，因 CC BY-NC 不在 SPDX 常见许可标识内；Copyright (c) 2026 dontbesilent）Commit 8b8e33f1ec

Control levels describe skill capabilities and suggested control measures; they do not represent quality or safety certifications, nor do they dictate usability. Usage decisions rest entirely with users and administrators. Different agent hosts apply varying constraints; do not assume hosts provide automatic protection. Third-party skill copyrights belong to their original authors. Use at your own risk.

1How It Works & Mechanism

A pure-prompt 'goal audit' skill: Determines whether a statement can guide action and identify completion, asking follow-up questions only when gaps affect deliverables, paths, or acceptance.

skills/dbs-goal/SKILL.md

你是 dontbesilent 的目标审计 AI。

The entire skill consists of questioning discipline and a goal card template; there are no scripts in the directory.

The value proposition is informational increment: Directly extract what has been stated, label safe inferences as assumptions, and only ask about information that will alter the outcome.

skills/dbs-goal/SKILL.md

用户已经说过的内容直接提取；能够安全推断的内容明确标注；只有缺口会改变交付物、执行路径或验收结果时才追问。

This is more fundamental than the Phase structure: It transforms 'goal clarification' from a questionnaire-style interrogation into minimal intervention.

The criterion is derived from 'meaning is use': The meaning of a term depends on the difference it produces in the current task.

skills/dbs-goal/SKILL.md

一个词的意义要看它在当前任务中产生了什么差异。

Goal phrasing must accomplish two things: make the next step certain, and make the completion state identifiable.

Explicitly rejects the grammatical word-deletion test in favor of semantic work detection (whether it alters deliverables, audience, scope, execution path, or acceptance).

skills/dbs-goal/SKILL.md

不要使用单纯的「删掉后句子是否仍然通顺」作为判据。

Counterexample provided: In 'deliver a CSV-formatted data table,' removing 'CSV-formatted' still leaves it grammatically fluent, but the delivery specification changes; thus, the term is doing work.

Uses five family resemblance characteristics (referentiality, falsifiability, completion state, grammatical soundness, embedded in context) to judge maturity, setting no rigid universal threshold and allowing clearance if three are met.

skills/dbs-goal/SKILL.md

用五条特征识别目标成熟度，不把它们设成统一硬门槛：

Supported by the source text: 'Usually satisfying three of them is sufficient to form a usable goal'—preferring clearance over perfection.

Four-path routing: Already actionable (clear directly) / Blocking gap exists (ask only 1 at a time) / Still a desire / Goal is merely a means.

skills/dbs-goal/SKILL.md

#### 路径 A：已经可以执行

Path A explicitly requires 'Do not continue asking questions just to finish the process.'; Path B specifies 'Ask only 1 question at a time and re-evaluate after receiving the answer.'

Four evaluation criteria for blocking gaps (different answers → different deliverables / different paths / different acceptance / unauthorized assumptions → rework risk), requiring questions to target specific decisions.

skills/dbs-goal/SKILL.md

> 这份报告只使用公开数据，还是也包含内部经营数据？两个范围会直接改变取数方式和可发布范围。

Example questions come with explanations of 'what the two scopes will directly alter,' letting users know why they need to answer.

Conditional tests are invoked based on gaps rather than a full questionnaire: referentiality, completion boundary, failure boundary, semantic work, context compatibility.

skills/dbs-goal/SKILL.md

### 语义做工测试

'Words without observable differences are marked as "not yet doing work." Words with differences but unclear boundaries are marked as "needs scoping."'—avoids judging idle terms merely by a word list.

Preserves user sovereignty: Can offer low-risk candidate judgments, but must not decide audiences, numbers, deadlines, purposes, or success criteria on behalf of the user.

skills/dbs-goal/SKILL.md

> 我先按「只使用公开数据」执行。如果你希望加入内部数据，再告诉我即可。

Candidate judgments must satisfy 'low risk, easy to correct, and will not conceal critical divergences.'

Outputs across three tiered templates based on sufficiency: Goal Card / Single-gap follow-up / Vague goal completion audit, strictly prohibiting verbose question-by-question logs.

skills/dbs-goal/SKILL.md

### 已经清楚：目标卡

The body text explicitly states: 'Unless the user requests a full report, do not output verbose question-by-question audit records.'

2Core Capabilities

01Extract seven goal card fields from context (Deliverable, Target Audience, Scope, Specifications, Completion Conditions, Constraints, Subsequent Use)

02Four-path routing (Clear / Blocking Gap / Desire / Means)

03Semantic work detection: Determine whether vague terms alter execution or acceptance

04Maturity rating based on five family resemblance characteristics

05Single-gap questioning discipline (Ask only 1 question at a time, first clarifying what is already established)

06On-demand invocation of conditional tests (Referentiality, Completion boundary, Failure boundary, Context compatibility)

07Three-tier output templates: Goal Card / Follow-up / Full Audit

08List of prohibitions (No repetitive questioning, no fixed questionnaires, no substituting user intent with arbitrary numbers)

4Recommended Governance & Evidence

Suggested Controls for Safe Operation

For Agent Runtime

Can be auto-loaded

For Administrators

Evidence & Triggers (File Paths & Excerpts)

SKILL.md: Pure prompt instructions and rule text; no script execution or external impact

Audit Notes: Evaluation target is the skill's own directory (dbs-goal/SKILL.md + agents/openai.yaml): zero scripts, zero network endpoints, zero credential reading, zero disk writes (dangerous token scan: ZERO HITS); all behaviors occur exclusively on the model output side. Points of attention: candidate assumptions may silently solidify, and audit conclusions are model-generated rather than code-enforced. Host tool and external platform behaviors are not used as criteria for rating.

5Second-Pass Verification

[ok] 纯 prompt、无脚本 — 目录仅 SKILL.md(10015B) + agents/openai.yaml(258B)；危险 token 扫描零命中。
[ok] 语义做工测试与 CSV 反例 — 原文含「不要使用单纯的「删掉后句子是否仍然通顺」作为判据。」与 CSV 反例整句，逐字命中。
[ok] 五特征与三项放行 — 正文列出可指物性/可否证性/有完成态/语法健全/嵌在上下文里，并写「通常满足其中三项即可形成可用目标。」
[ok] 四路径与单缺口提问 — 路径 A/B/C/D 四节均在原文；「一次只问 1 个问题，等待回答后重新判断。」逐字命中。
[ok] 三档输出模板 — 「### 已经清楚：目标卡」「### 仍有一个阻塞性缺口」「### 模糊目标完成审计」三节均在原文。
[ok] 外部依赖为空 — 无 URL/CLI/包依赖；无 process.env/密钥读取；无写盘指令。
[ok] description 与内容相符 — description 的三项承诺（提取已有约束、只追问影响执行或验收的信息、整理成可行动可检查的交付物）分别对应提取表、提问纪律、目标卡模板。
[ok] commit / license / 元数据 — HEAD = 8b8e33f1ecaed8cee606fe950c4426b525ead314；LICENSE = CC BY-NC 4.0（NonCommercial）；stars 10121、pushed_at 2026-09-07T03:48:59Z。

6Verdict & Scenarios

Minimal intervention: Inquires only when gaps alter deliverables, execution paths, or acceptance criteria; directly clears clear goals.

Reusable and falsifiable criteria: Semantic work detection provides counterexamples to prevent misjudgments based solely on grammatical fluency.

No rigid universal threshold: Clears when three of the five characteristics are met, reducing process overhead.

Preserves user sovereignty: Only provides low-risk candidate judgments and forbids inventing metrics or success criteria on the user's behalf.

Sufficiency-driven output: Chooses from three templates based on clarity level, avoiding verbose question-by-question logs.

Best for: Suitable for 'vague tasks leading to repeated rework' scenarios: transforms ambiguous requests (e.g., 'make a report,' 'build a personal brand,' 'optimize this') into goal cards with clear deliverables, identifiable completion conditions, and concrete next steps; also suitable as an acceptance criteria check prior to delivery.

Not for: Not suitable for execution-oriented tasks that already have precise specifications (will be routed to Path A and directly cleared, offering limited value); not suitable for users expecting the skill to unilaterally invent numbers, deadlines, or success criteria—it preserves user sovereignty by design; not suitable for tasks requiring fact-checking, data research, or document-based record-keeping (this skill only produces in-chat text and does not write files).

Install Agent-Ready

① Verified Mirror 2026-09-15

Option A · Direct DownloadDownload dbs-goal.tar.gz

sha256: 449a53c1d3bb4603…

Option B · Copy Guide for Agent

install.json

Agents automatically download and verify mirrors from this JSON guide.

② Upstream GitHub

GitHub Repo ↗

Pinned to commit 8b8e33f1ec.