dotharness/dotcraft/dotcraft-report-issue
Claude Code and agent skill by dotharness/dotcraft
- community
Checking that the SKILL.md file is still in the repository.
Turn a DotCraft failure or diagnosis into a clear, well-structured GitHub issue for the DotHarness/dotcraft repository, and produce a prefilled "new issue" URL the user can review and submit. Use after dotcraft-error-diagnosis (or when a user wants to report a bug), to draft the issue title and body and hand off a ready-to-open link. Does not submit on the user's behalf.
Install
npx vskill@latest install dotharness/dotcraft/dotcraft-report-issueSKILL.md
DotCraft Issue Report
Overview
Use this skill to convert the output of dotcraft-error-diagnosis or a user's bug description into a single, high-signal GitHub issue for the DotHarness/dotcraft repository. The goal is a concise report a maintainer can triage: clear title, public reproduction context, non-sensitive diagnostics, and environment.
This skill drafts and hands off; it never submits. The user always reviews the draft and clicks submit on GitHub themselves.
Important distinction: thread IDs and local DotCraft evidence are useful for local diagnosis, but they are not automatically suitable for the public GitHub issue. Treat any source thread ID supplied by Desktop as a local lookup hint unless the user explicitly approves sharing it.
Safety And Privacy Rules
- Never include secrets: API keys, tokens, full auth headers, cookies, or absolute paths that reveal usernames you were not asked to share. Redact to
<redacted>. - Keep issue drafts minimal by default. Include the problem summary, impact, environment, short error preview, public reproduction steps, and necessary non-sensitive diagnostic conclusions.
- Thread IDs, rollout paths, trace/session IDs, user prompts, model output, and conversation excerpts may be used as local diagnosis inputs only. Do not copy them into the issue title or body by default.
- Prefer component names, tool/provider names, error classes, finish reasons, status codes, timestamps without identifying IDs, and short sanitized error previews over internal IDs or conversation content.
- Convert operational-log evidence into a sanitized finding. Do not include raw log lines, absolute local paths, usernames, process or connection IDs by default.
- Keep user prompts and model output out of the issue unless the user confirms they are safe to share and they are needed to reproduce.
- If the issue depends on a specific conversation operation, ask before including any thread/conversation information. Explain which fields would be included, why they matter, and how they will be redacted.
- If the user does not approve sharing thread/conversation details, replace them with a sanitized summary such as
a resumed Desktop thread after compaction. - Show the user the full draft before producing the submit link. Make it easy to edit title and body.
Inputs To Gather
- What failed — one sentence (from the diagnosis finding, or the user).
- Diagnostics — optional, concise bullets: component/tool/provider, error class, finish reason, status code, sanitized log finding, short error preview, and version/model/provider details. Use local IDs only to diagnose, not as default issue content.
- Reproduction — the minimal public steps that triggered it, if known. If they are not known, write
not yet knownor ask the user for publicly shareable steps. - Environment — capture automatically where possible:
- DotCraft version (and Desktop version if applicable)
- OS and version
- Active model / provider and reasoning effort
- Runtime (.NET version) when relevant
When To Ask Before Sharing Conversation Details
Ask the user for permission before adding thread IDs, rollout paths, trace/session IDs, user prompts, model output, or conversation excerpts to the issue. Only consider including them when the bug itself depends on conversation-specific operations, for example:
- resume, fork, compaction, or context restoration behavior
- thread binding, queued input, or cross-thread routing
- a specific sequence of tool calls or approvals
- trace/session persistence mismatches that cannot be described without the affected session identity
When asking, name the exact fields you want to include and propose the shortest redacted form. If the user does not confirm, keep the issue public-safe and describe the scenario generically.
Issue Format
Title: a specific, searchable one-liner. Pattern: <area>: <symptom> when <trigger>. Examples: Edit tool: "old_string not found" after compaction, Provider: 400 unsupported param on stream with reasoning=high.
Body (Markdown):
Handoff
After the user approves the draft, produce a prefilled new-issue URL so they can open it in the browser and submit:
- URL-encode both the title and the body (encode spaces, newlines,
#,&, etc.). - GitHub caps the URL length; if the encoded body is very long, trim the lowest-value details (long evidence lists) and tell the user the full draft is above so they can paste it if needed.
- Present the prefilled URL along with the raw title and body so the user can open it or paste it wherever they prefer.
- Remind the user to review for anything sensitive before submitting.
Composing With dotcraft-error-diagnosis
Typical flow: dotcraft-error-diagnosis runs first and returns Finding / Evidence / Fix / Residual risk. Map that directly:
- Finding → Summary + Diagnosis.Finding
- Evidence → Diagnosis.Diagnostics after summarizing log evidence and removing raw log lines, local paths, usernames, process/connection IDs, thread IDs, rollout paths, trace/session IDs, prompts, model output, and conversation excerpts.
- Fix → omit from the issue (that is the maintainer's call), or include only if the user wants to propose it.
- Residual risk → Notes.
Security
- Security scanPassed, 90 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingdesktop/resources/plugins/dotcraft-bundled/plugins/dotcraft/skills/dotcraft-report-issue/SKILL.md
Popularity
- 7-day trend score
- 0
Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/dotharness/dotcraft/dotcraft-report-issue)