Skip to content

Latest commit

 

History

History
77 lines (57 loc) · 3.41 KB

File metadata and controls

77 lines (57 loc) · 3.41 KB

Agent Notes

This project is running in an exe.dev VM. Use only documented exe.dev features:

Undocumented local endpoints are internal infrastructure and should not be used.

Project Work Tracking

Always use this repo's own Tala skill as the durable project work ledger.

Harness-agnostic skill

  • Skill: .opencode/skills/tala-project-tracker/
  • Database: .tala/tala.db under the project root
  • MCP server: go run ./cmd/tala-mcp-stdio -db .tala/tala.db
  • HTTP server URL for browser workflows: http://127.0.0.1:8081
  • HTTP server command: go run ./cmd/tala -addr 127.0.0.1:8081 -db .tala/tala.db

Codex plugin (Codex-only)

  • Plugin: plugins/tala-project-tracker
  • Skill: tala-project-tracker from the plugin
  • Codex MCP launcher: plugins/tala-project-tracker/scripts/tala-mcp-stdio.sh
  • Codex web server launcher: plugins/tala-project-tracker/scripts/tala-web-server.sh

Use the Tala skill for planning, implementation tracking, progress updates, and handoff notes. Before reading .tala/tala.db directly, check for the repo-local Tala MCP tools. If the MCP tools are not already visible, use tool discovery/search for "tala issue search create comment update image upload project tracker MCP" with enough result slots to expose the full tool set, then use the available tools. Expected tools include issue_search, issue_get, issue_create, issue_update, issue_comment, image_upload, and relationship/status helpers such as issue_set_status, issue_set_priority, issue_set_parent, issue_add_blocker, and issue_remove_blocker. In OpenCode these tools are typically prefixed as mcp__tala.<tool>; other MCP clients may use a different prefix, but the underlying tool names and arguments are the same. Only read .tala/tala.db directly for diagnostics or verification when MCP tools are unavailable. Do not use another Tala database unless the user explicitly requests it.

Before implementing an accepted plan, ask the user whether to create a new Tala issue or reuse an existing issue for the work. If they want tracking, search first, create a new issue with the issue_create tool only when no clear match exists, mark the active issue in_progress, and add progress, completion, or handoff notes with the issue_comment tool.

Before finishing any work that creates temporary issues, tags, comments, or other test records in this repo's .tala/tala.db, clean them up or document why they must remain. This applies especially to browser smoke tests that create records with names such as browser_smoke_*, detail-smoke-*, profile-smoke-*, overflow-*, or other run-specific markers. Run a post-check against .tala/tala.db to confirm temporary entries are gone, while preserving real project issues, roadmap items, and durable bug/task records.

Product Design References

The technical design document is the source of truth for the issue tracker:

  • docs/issue-tracker-design.md

The Google Stitch UI/UX mockups and design-system assets live under:

  • .stitch/DESIGN.md
  • .stitch/metadata.json
  • .stitch/designs/

Stitch project:

  • Project title: Tala Issue Tracker Mobile Mockups
  • Project ID: 2814771326082574657
  • Design system asset: assets/dfb1f957bbf24691aa401e82e1a35e7d

When implementing frontend views, align behavior with docs/issue-tracker-design.md and align visual/UI decisions with the Stitch mockups and .stitch/DESIGN.md.