openclaw-agent-finn

Summary

Finn provides a QA Analyst workspace for OpenClaw venture projects. It enables methodical, unforgiving product quality assurance for live proof-of-concept (POC) releases—prioritizing P0 criteria and real user experience. Finn’s purpose is to catch critical issues before projects proceed, reporting actionable findings in a rigorous, evidence-driven format.

Included Assets

• AGENTS.md — Finn’s role/rules reference and QA protocols
• SOUL.md — Core principles, agent identity, and non-negotiables
• TOOLS.md — Setup for GitHub, browser, shared ventures vault, and workspace memory
• agents/finn.md — Step-by-step QA instructions for typical cycles
• agents/finn-ref.md — QA report and feedback templates (PASS/FAIL)
• memory/tasks.md — Active QA tasks tracker
• DISCORD.md — IDs and channel references for workspace Discord integration
• HEARTBEAT.md — Latest status and operational checks for Finn
• IDENTITY.md — Summary of Finn’s character and role in the venture studio
• REFLECTION.md — Weekly QA review and update prompts
• USER.md — Profile of primary user (Lukas), including communication style and priorities

How to Use

1. Trigger a QA Cycle

   • Set a project TICKET.md stage to qa-briefed to assign work to Finn.

2. Follow Testing Protocol

   • Finn reacts to new QA tasks and consults:
     • Sprint contract for acceptance criteria
     • Build report and live URL
     • Design spec for intended user experience
   • All QA is conducted by interacting with the live POC (not code).

3. Quality Assurance Process

   • Follow defined procedure in agents/finn.md, focusing on:
     • P0 criteria first (must pass)
     • P1 criteria (tracked, but not blocking)
     • Universal P0 defaults (e.g., no 500/404 errors, core flows work)
   • Mark progress in Discord with specific emojis (see AGENTS.md).
   • Screenshot and document all key steps.

4. Report & Verdict

   • Complete a QA report or feedback using the templates in agents/finn-ref.md.
   • Record observations, step-by-step evidence, and explicit PASS/FAIL verdicts.
   • Save results in the appropriate venture’s /build/ directory.

5. Weekly Reflection & Maintenance

   • Use REFLECTION.md to review recent QA outcomes, update rules, and document lessons learned.

Notes

• All QA must occur on the live app—reviewing code or freestyle testing is not permitted.
• No filler or affirmations: Reports are concise, direct, and actionable.
• Discord integration: Use workspace-specific agent accounts and designated channels per DISCORD.md.
• Agent memory: QA logs and context are maintained locally; agent wakes up fresh for each new session.
• Identity: Finn acts with skepticism, prioritizes broken experience over partial success, and escalates immediately when evidence is weak.

For detailed working instructions and report formats, see agents/finn.md and agents/finn-ref.md.