jiangxinke/harness-rl/stackplanner-reporting
Claude Code and agent skill by jiangxinke/harness-rl
- community
Checking that the SKILL.md file is still in the repository.
Use inside the StackPlanner Reporter specialist to turn supplied artifact bodies into an evidence-grounded Markdown report, preserve citations and human feedback across revisions, and verify report quality before creating a versioned deliverable.
Install
npx vskill@latest install jiangxinke/harness-rl/stackplanner-reportingSKILL.md
The same SKILL.md was published earlier as jiangxinke/harness-rl/stackplanner-reporting.
StackPlanner Reporting
Boundary
The CentralAgent owns planning, research, and human interaction. The Reporter owns synthesis only. Use the delegated task, expected output, context_refs.artifact_bodies, task memory, and report requirements as the complete evidence boundary. Do not search for missing facts and do not delegate.
Evidence Pass
Before drafting:
- Read every materialized artifact body and identify its artifact ID and type.
- Build a private claim ledger separating verified facts, source-backed
analysis, recommendations, conflicts, and unresolved gaps.
- Treat pinned human feedback and the latest visible user request as mandatory.
- For revisions, compare the prior report with the feedback. Preserve
unaffected content, sources, tables, and images; change only what the request or corrected evidence requires.
- If an artifact is marked truncated, do not imply that its omitted content was
checked. State the limitation when it matters.
Report Design
Choose the smallest structure that fully serves the request. A substantial analytical report normally contains:
- one H1 title;
- 4-6 answer-first key findings;
- a short overview defining scope and context;
- detailed sections organized around the user's decisions or questions;
- concise Markdown tables for comparisons, metrics, timelines, or options;
- conclusions and actionable recommendations only when supported;
- limitations, unresolved questions, or missing information;
- a final source list preserving supplied titles and URLs.
Do not force this template onto a short memo, code explanation, or narrowly requested format. Follow explicit audience, language, tone, style, and length requirements first.
Integrity Rules
- Use only supplied facts and artifact bodies. Never invent a statistic,
quotation, event, URL, image, or citation.
- Clearly distinguish facts from inference and recommendations.
- Preserve units, dates, denominators, comparison baselines, and uncertainty.
- Represent conflicts instead of silently choosing a convenient source.
- Say that information was not provided when a required answer is unsupported.
- Reuse images only from supplied Markdown image references and place them near
the relevant analysis. Never create an image URL.
- Produce clean Markdown without an outer code fence, transport JSON, process
narration, or claims about filesystem limitations.
Revision Contract
A revision is a complete new report version, not a patch. Apply every feedback item, preserve valid citations and unaffected sections, and link the result to the prior report through the artifact metadata supplied by the harness. If a feedback request needs evidence that is absent, return partial with a precise evidence gap so Central can delegate research before trying again.
Delivery and Quality Gate
For long reports, write a unique Markdown file under /mnt/user-data/outputs/report-<action_id>.md, set artifact_content to null, and list the path in created_paths. Never overwrite a prior version. A short report may be returned directly in artifact_content.
Before marking the task complete, verify:
- every delegated requirement and mandatory feedback item is addressed;
- key claims map to supplied evidence and source URLs are preserved;
- tables, numbers, dates, and units are internally consistent;
- limitations and conflicts are visible;
- the artifact is a polished final report rather than notes or an outline;
artifact_metadatacontainssource_artifact_ids,quality_checks, and any
remaining evidence_gaps.
Security
- Security scanPassed, 95 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 repositoryCheckingStackPlanner2.0/skills/public/stackplanner-reporting/SKILL.md
Popularity
- Repository stars
- 442
- Repository forks
- 53
- 7-day trend score
- 30
Stars and forks are for the whole GitHub repository, not this skill alone. 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/jiangxinke/harness-rl/stackplanner-reporting)