[pranshuchittora]

Why

Projects using agent-qa need guidance for authoring journeys and interpreting recorded evidence without treating setup checks or missing runs as a pass. Add a Katara-owned skill for agent-qa by vostride, with natural-language web/mobile testing, advanced memidex memory, and self-improvement.

What changed

• Add a focused skill with project/run and evidence/memory references checked against agent-qa 0.2.3.
• Register Katara ownership, add the request route, and regenerate the roster tables.
• Preserve explicit non-execution requests and require evidence before reporting outcomes or applying knowledge corrections.

Scope

• Covers agent-qa workflows in projects that use it; runtime installation and model setup remain project choices.
• No new hooks, scripts, bundled runtime, or automatic MCP configuration.

Blast Radius

Adds one optional skill and its routing. Existing skills and runtime configuration retain their behavior.

Verification

• Native validator reports 59 skills and 9 agents; generator drift check, all 253 Bun tests, type check, and diff check pass. Test fixtures use process-local Git signing disabled.
• Claude plugin validation succeeds; isolated Skills CLI installs for Claude Code and Codex preserve all three skill files.
• Two model trials used supported record readers and correctly reported missing completed-run evidence; compared project files stayed unchanged. A blinded review supports that narrow result. Full pai-sho validation remains incomplete: QA terminology prevents strict token exclusion, model-family diversity is limited, and scoped raw transcripts are unavailable. Live host trigger activation and application journeys were not exercised.

[Copilot]

🟢 Approval recommended

The optional guidance, registration, and routing changes are consistent, with no identified blocking issues or runtime configuration changes.

No open issues 🎉

What changed in this PR

Adds an optional Katara-owned agent-qa skill for authoring journeys, running requested tests, and reviewing recorded evidence without changing runtime configuration.

Changes:

• Documents project setup, execution, evidence readers, and authorized memory corrections.
• Distinguishes preparation checks from completed runs and preserves explicit non-execution requests.
• Registers the skill and updates Katara’s routing and generated rosters.

File	Description
plugins/​team-avatar/​team.json	Registers agent-qa under Katara.
plugins/​team-avatar/​skills/​avatar-mode/​SKILL.md	Adds request routing and updates the roster.
plugins/​team-avatar/​skills/​agent-qa/​SKILL.md	Defines the workflow and evidence safeguards.
plugins/​team-avatar/​skills/​agent-qa/​references/​project-and-runs.md	Documents setup, authoring, execution, and optional MCP configuration.
plugins/​team-avatar/​skills/​agent-qa/​references/​evidence-and-memory.md	Documents evidence inspection and authorized knowledge corrections.
plugins/​team-avatar/​README.md	Updates Katara’s skill roster.
plugins/​team-avatar/​agents/​katara.md	Adds the skill-loading instruction and roster entry.

_{Review effort: Balanced}