The server defines 23 tools across issues, projects, users, and milestones with generally complete JSON Schema input definitions and descriptions. Strengths: all tools have descriptions (10-250+ chars), all parameters have type definitions, many have enums (priority, visibility, status), and composite tools provide good chaining support (e.g., createIssue accepts project_key OR project_id). Weaknesses: descriptions vary significantly in quality and LLM-optimization; several tools lack guidance on when to use them vs similar tools; output schemas are not explicitly documented; error handling descriptions are minimal; some parameter descriptions are generic or lack validation rules; no guidance on pagination limits or result caps for list operations.
Tools (26)
kiket_create_anchor_proofwriteauthsource verified
Create a local or submitted audit anchor proof for report, evidence, or finding integrity through the authorized platform API.
Pagination guidelines missing. listIssues, listProjects, listUsers, listMilestones accept page/per_page but lack stated result caps or guidance on what 'reasonable limit' means (typically 20-50 items).
listIssueslistProjectslistUserslistMilestones
Recommendations
Document output schemas for all 23 tools. For listIssues, specify: returns array of {id, key, title, status, assignee_id, labels, priority, parent_id, created_at, updated_at}. For getIssue, same structure but single object. For createIssue, return the created issue object with all fields. This enables LLMs to chain calls (e.g., create issue → add comment using returned issue ID).
Add explicit result caps to pagination-aware tools. E.g., 'listIssues returns max 100 items per call; use page/per_page to iterate. Default limit: 20 items.' Include guidance: 'For most use cases, per_page=20 is sufficient; increase only if you need detailed batch processing.'
For destructive tools (deleteProject, deleteMilestone), add a dry-run pattern or confirmation step. E.g., 'deleteProject requires confirmation. First call with dry_run=true to see what will be deleted, then call with dry_run=false to confirm.' Or document: 'Deletes the project and all associated issues. This is irreversible.'
For transitionIssue, add an enum constraint or discovery tool. Best approach: call getIssueSchema first to see available transitions. Update description: 'Use getIssueSchema(project_key) to discover valid transition keys for the workflow before calling transitionIssue.'
Enhance descriptions for list tools to disambiguate. E.g., 'listIssues: Fetch issues within a project, optionally filtered. Use for bulk operations or multi-issue views. For a single issue, use getIssue.' 'listMilestones: Fetch release/milestone metadata for a project. Use to understand which milestones exist before assigning issues.'
Score history
Overall score trend
↓ 50 points across a rubric change (v1 → v2)
0/100
Scored
Grade
Overall
Spec posture
Rubric
2026-09-22
F
0
2026-07-28+
v2
2026-03-09
D
50
2024-11-05+
v1
Get the evidence provenance graph for a scoped investigation (case, process, finding, or workspace).
Descriptions for deleteProject and deleteMilestone (65 chars each) lack guidance on what happens to related data or recovery options. No confirmation/dry-run pattern present.
Parameter 'transition' in transitionIssue lacks enum values or guidance on valid transitions. LLMs must call getIssueSchema first or guess valid values, causing wasted calls.
When multiple tools access similar resources (e.g., listIssues vs listMilestones), descriptions don't clearly distinguish when to use each. Risk of LLM confusion.
Descriptions for list tools (listProjects, listUsers, listComments, listMilestones) are generic (65-75 chars) and lack context on what 'visible to the authenticated user' means or permissions required.
listProjectslistUserslistCommentslistMilestones
Add permission/scope hints to all tools. E.g., createIssue: 'Requires write permission on the project. If you get a 403 error, ask the user if they have project write access.' deleteProject: 'Requires admin permission on the project.' This guides LLM error recovery.
For custom_fields in createIssue/updateIssue, add guidance: 'Custom field keys and valid values are project-specific. Always call getIssueSchema(project_key) first to see available fields.' This prevents hallucinated field names.
For listComments, add: 'Returns comments in chronological order. To check if an issue has been resolved with comments, list comments and check timestamps/authors.' This guides LLM reasoning about when to use the tool.
For createProject, explicitly state validation rules in description: 'name: 2-100 chars. project_key: Uppercase alphanumeric (regex ^[A-Z][A-Z0-9]*$, 2-10 chars recommended). github_repo_url: Must be a valid HTTPS GitHub URL.' LLMs often miss regex patterns in JSON Schema.
Add error guidance to all WRITE and DESTRUCTIVE tools. E.g., 'If you get status 422 (ValidationError), the description will say which field failed and why. Adjust that field and retry. If status 401, authenticate first.' This turns raw errors into recovery paths.