| name | mcp-java-dev-tools-issue-report |
| description | Create sanitized, reproducible bug/issue reports from session evidence. Use for issue tickets, failure summaries, repro steps, exceptions, or failing endpoints. |
MCP Java Dev Tools Issue Report
Create a reproducible issue report from the active session. Prefer current-thread evidence over speculation.
Contract
- Use this skill for issue reporting only.
- Treat the current session as the primary source of truth.
- Gather only the minimum details needed for a developer to reproduce and investigate the issue.
- Never emit raw secrets, enterprise identifiers, or proprietary class/package names in the final output.
- If a fact is missing, say it is missing. Do not invent repro details.
Workflow
- Read the user request and identify the reporting target:
- failing method
- failing endpoint
- failing workflow
- runtime/probe failure
- Mine the current session and workspace for evidence:
- user-provided symptoms
- HTTP requests and responses
- stack traces and logs
- probe results and captures
- relevant source locations only when needed to explain the issue
- Normalize the issue into developer-useful facts:
- observed behavior
- expected behavior
- reproducible trigger
- affected component
- impact
- Sanitize the output:
- redact secrets
- anonymize identifiers
- rewrite package names to neutral namespaces
- preserve technical role and structure
- Perform a final leak review before answering.
Evidence Rules
- Prefer direct runtime evidence over static guesses.
- Prefer exact trigger requests over abstract descriptions.
- Use short excerpts or summaries instead of long raw dumps.
- When probe evidence exists, use it to support the issue but do not make MCP internals the center of the report.
- If the issue was not fully reproduced, state that clearly and separate:
- confirmed facts
- likely cause
- missing evidence
Sanitization Rules
Read references/sanitization-rules.md before generating the final report.
Apply the rules in that file to:
- class names
- package names
- organization names
- service names
- hostnames and URLs
- credentials and auth material
- business-sensitive terminology
Output Format
Read references/output-template.md and follow it exactly.
Required properties of the final report:
- Short and reproducible.
- Safe to share with developers without leaking enterprise details.
- Preserve technical meaning after anonymization.
- Keep naming consistent across the whole report by using a stable alias map.
Reporting Rules
Issue Title must be a single concise sentence.
Issue Description must summarize the failure, trigger, and effect.
- Supporting sections must contain only evidence that materially helps reproduction or investigation.
- Include exact request shape when known:
- method
- sanitized URL/path
- sanitized query params
- sanitized headers
- sanitized body
- Include expected behavior only when it can be inferred from the session or user statement.
- Include impact only when it is grounded in evidence.
Anonymization Policy
Preserve architectural meaning. Remove proprietary identity.
Good:
CompanyClassA -> ClassA
SynonymsRuleController -> RuleController
AcmeCatalogService -> CatalogService
com.company.catalog.web.controller -> com.example.catalog.web.controller
Bad:
SynonymsRuleController -> <REDACTED>
CatalogService -> Thing
- removing all class/package information
Final Leak Check
Before answering, verify the report does not contain:
- bearer tokens
- cookies
- passwords
- API keys
- internal hostnames
- company-specific package roots
- proprietary class or service names when a generic alias can preserve meaning
If leakage remains, rewrite before answering.