Home / All Skills / Docs & Knowledge / dbs-report
English (EN)简体中文 (ZH)日本語 (JA)
Docs & Knowledge · dontbesilent2025/dbskill

dbs-report

Merges multiple dbs-save archives into a deliverable Markdown report. Use when the user requests summarizing diagnoses, compiling reports, or generating shareable materials.

L3 Networked · Declare domain & key scopeV1 Commit-PinnedNetwork AccessWrite FileDual-pass Audit
Author dontbesilent2025GitHub dontbesilent2025/dbskill ↗Stars 10121License CC BY-NC 4.0（仓库根 LICENSE：Attribution-NonCommercial 4.0 International (CC BY-NC 4.0)；署名 + 禁止商业使用；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-orchestrated "artifact generator": inputs are archived markdown files saved to disk by dbs-save, and the output is a newly timestamped report markdown file; the entire process only reads, merges, categorizes, and transfers data.

skills/dbs-report/SKILL.md
报告内容必须来自 `{存档根目录}/sessions/{项目名}/` 下的存档文件。
The repository README.md positions it under "Record and review long-term decisions | /dbs-decision, /dbs-save, /dbs-restore, /dbs-report".

Shares the same archive root directory resolution rules with dbs-save/dbs-restore (default/project/custom modes in .dbs/config.json), and all three only read configurations without writing them.

skills/dbs-report/SKILL.md
`dbs-report` 只读取配置，不修改配置。
If resolution fails or paths point to /, user home, or the project root, it halts report generation and explains the configuration issue, rather than silently falling back to default locations.

Credibility relies on "traceability" rather than "good summarization": the primary core principle forbids summarizing out of thin air from the conversation; every detail must point back to specific fields in specific archives.

skills/dbs-report/SKILL.md
**不从对话凭空总结**。报告内容必须能追溯到具体存档文件的具体字段。这是报告的可信度
Therefore, five of the report's six sections are transfers + deduplication + categorization, with only the "Evolution of User's Chief Complaint" section allowing brief summarization.

Multiple runs never overwrite files, using timestamps to provide "diagnostic snapshot comparison" capabilities.

skills/dbs-report/SKILL.md
每次新生成一份，**永不覆盖**。文件名带时间戳，方便对比不同时点的诊断快照。
Path rule: {archive_root}/reports/{project_slug}/{YYYYMMDD-HHMMSS}-{project_slug}.md; runs mkdir -p first if the directory does not exist.

Does not arbitrate conflicting conclusions: lists both newer and older versions, placing the newer one first and annotating the older one as "(Amended by subsequent diagnosis)".

skills/dbs-report/SKILL.md
如果一条结论在后续存档里被推翻或修正，**两条都列出来**，新的在前，旧的在后并加 `（已被后续诊断修正）` 标注。
"Conflicts in conclusions across multiple archives → list all of them" is explicitly restated in edge cases.

Deliberately avoids data masking: sensitive figures and customer names in archives appear as-is in the report, pushing sanitization responsibility back to the dbs-save stage.

skills/dbs-report/SKILL.md
存档文件里有用户标的「敏感信息」（比如真实收入、客户名字）→ 报告原样保留。不做脱敏。
The skill's rationale is that "this was stored by the user themselves," and it also constrains output formats: it does not actively generate PDF/HTML.

Strict minimum data threshold: rejects 0 archives directly, advises against 1 archive and asks whether to force merge, and only enters merging with ≥2 archives (the same criteria apply after --since filtering).

skills/dbs-report/SKILL.md
0 个文件 → 「`{项目名}` 下还没有可汇总的存档。
Prevents generating uninformative empty reports or pseudo-merged outputs from single files.

Integration with dbs-wechat-html is conditional: only prompts once if the user genuinely needs to publish, remaining unprompted otherwise.

skills/dbs-report/SKILL.md
想发公众号或群里，可以用 `/dbs-wechat-html` 把这份 markdown 转成微信后台粘贴版。
Original precondition: "If the user needs to publish the report to an Official Account or group chat, add one line"; postcondition: "Do not add if the user has no publication needs".

2Core Capabilities

01Multi-archive merging: reads in chronological order by YYYYMMDD-HHMMSS, parsing frontmatter and the six body sections
02Six-section report structure: Chief Complaint Evolution / Confirmed Conclusions / Rejected Directions / Current Unresolved Issues / Recommended Next Steps / Archive Index Table
03Conclusion deduplication and amendment annotation (merges semantically similar items, sorted in reverse chronological order, keeps both overturned conclusions)
04Date filtering via --since and project switching via --slug
05Unresolved issues aggregation: merges unverified hypotheses from in-progress archives + directions proposed in the earliest archives that were never addressed
06Next steps prioritization (prioritizes latest / recurring but untaken / marks superseded items)
07Localized Chinese display of status fields (in-progress/resolved/abandoned → 进行中/已结论/已放弃)
08Does not search private files across directories: operates strictly within the sessions directory pointed to by the current configuration

4Recommended Governance & Evidence

Suggested Controls for Safe Operation
For Agent Runtime
Declare allowed domains and key scope
For Administrators
Least privilege keys, retain request logs
Evidence & Triggers (File Paths & Excerpts)
  • 环境变量 / API Key: Read vendor service credentials: Reads .dbs/config.json to resolve the archive root directory (mode=default/project/custom + root) — this is the user's own configuration
  • 本地工作目录: File read/write and artifact output: {archive_root}/reports/{project_slug}/{YYYYMMDD-HHMMSS}-{project_slug}.md (new reports, never overwritten; executes mkdir -p if the directory does not exist)
Audit Notes: The evaluation target is this skill's own directory: no scripts, no network endpoints, no credential reading, but SKILL.md explicitly instructs the agent to read local archives and write new report files ({archive_root}/reports/…md), falling under the "local read/write only, no external egress" tier. Key considerations focus on privacy rather than code execution: reports deliberately do not mask data (retaining revenues and client names as-is), and the output is "shareable material"—sharing the report exposes all sensitive details contained within the archives; the skill explicitly states this stance in its edge cases.

5Second-Pass Verification

  • [ok] 外部资源/工具：是否存在未声明的依赖或格式转换器 — 无 URL、无包、无 CLI 调用；PDF/HTML 相关语句只有禁止性的（不主动出 PDF / HTML / 其他格式）。external_deps 空数组成立。
  • [ok] 安全结论：写入面是否只有报告文件 — 全文写入只有 mkdir -p 与报告文件生成；明确不修改 .dbs/config.json（「只读取配置，不修改配置」），不删除任何存档，不改写原始 sessions 文件。
  • [ok] 实现原理是否成立：报告能否真正做到「每条可追溯」 — 六段中五段规定逐条携带来源（「· 出自 {对应存档的标题} · {对应日期}」）、附录给出文件名索引表，「结论冲突两条都列」避免丢失信息。机制可核查。
  • [ok] 功能声明 vs 实际能力是否存在夸大 — description 只说「合并成可交付的 Markdown 报告」，未承诺分析深度或结论质量；SKILL.md 亦把「总结」限制到一段。一致。
  • [unlocatable] 证据定位：仓库级元数据（stars/last_push）与 installs — stars=10121、last_push=2026-09-07 来自站点池 data/pool.json（GitHub API 采集），pin 本地副本无法逐字定位；installs 未采集记 null。license 可在仓库根 LICENSE 逐字命中。

6Verdict & Scenarios

  • Transforms "diagnostic logs" into "deliverables": six-section structure + index table + timestamped files enable the report to be shared directly with partners and external consultants or used for cross-month reviews
  • Credibility-first design: prohibits baseless summarization, includes provenance for each section, lists both conflicting sides, localizes status fields—outputs are ideal for archival reconciliation
  • Immutable artifacts: creates a new file each time and never overwrites, turning the reports themselves into a comparable time series
  • Shares configuration resolution rules with storage subsystems (dbs-save/dbs-restore), keeping paths and wording consistent so users only need to understand the "archive root directory" concept once
  • Restrained: does not actively change artifact formats (markdown only), does not proactively upsell adjacent skills (only mentions dbs-wechat-html once when publishing needs exist)
    • Best for: Suitable for users conducting long-term diagnostics on the same project who need to organize fragmented conclusions into external materials: reconciliation documents for partners/consultants/clients, periodic reviews, and tracking conclusion evolution across months. Forms a complete "save-resume-report" pipeline with dbs-save / dbs-restore.
    Not for: Not suitable for scenarios where users want the AI to "summarize what we discussed" without archives (the skill explicitly forbids baseless summarization and will reject such requests upfront); not suitable for workflows requiring external publication after data masking (it deliberately avoids masking); not suitable for delivery requirements needing PDF/HTML/PPT finished products (outputs markdown only and explicitly avoids proactive conversions); not suitable for scenarios where the user only wants to inspect a single archive (merging 1 archive will be discouraged).
    Install Agent-Ready
    ① Verified Mirror 2026-09-15
    Option A · Direct DownloadDownload dbs-report.tar.gz
    sha256: 1daf478be83e49ea…
    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.