/qa-catalog:run — Execute one task

/qa-catalog:run — Execute one task

Project context

• Task id: $ARGUMENTS
• Task file: !ls QA-tests/tasks/$ARGUMENTS*.md 2>/dev/null | head -1
• Catalog present: !test -f QA-tests/catalog.json && echo YES || echo NO

Runtime settings

Setting	Value
Dev URL	${user_config.dev_url}
Browser channel	${user_config.browser_channel}
Headless	${user_config.browser_headless}
Settle ms	${user_config.settle_ms}
Auth mode	${user_config.auth_mode}

Instructions

Locate the task file matching $ARGUMENTS. If none found, list available task ids and stop.

Phase 1 — Prepare run directory

1. Generate runId as a UTC ISO-8601 timestamp with : replaced by - (example: 2026-05-28T14-22-11Z).
2. Resolve runDir = QA-tests/results/runs/<runId>/<taskId> and create it (mkdir -p).
3. Write QA-tests/results/runs/<runId>/run.json:

   {
     "runId": "<runId>",
     "startedAt": "<ISO-8601>",
     "filter": "$ARGUMENTS",
     "settingsSnapshot": { ...effective user_config values... },
     "tasks": [{ "taskId": "<id>", "taskFile": "<path>", "route": "<route>" }]
   }
   

Phase 2 — Dispatch the runner

Resolve per-role credentials first. When auth_mode is per-role, load the credential map (secrets interpolated from env vars):

node "${CLAUDE_PLUGIN_ROOT}/scripts/auth-resolve.mjs" --json

Use the returned roles object as credentialsByRole. If the file is absent (present: false) or auth_mode is not per-role, omit credentialsByRole and let the runner fall back to the single shared credential below. Never print resolved passwords back to the user.

Spawn the project-level agent qa-test-runner with this payload:

{
  "taskId":   "<resolved id>",
  "taskFile": "<repo-relative task file>",
  "runDir":   "<runDir>",
  "runId":    "<runId>",
  "devUrl":   "${user_config.dev_url}",
  "settings": {
    "browserChannel": "${user_config.browser_channel}",
    "headless":       ${user_config.browser_headless},
    "settleMs":       ${user_config.settle_ms},
    "authMode":       "${user_config.auth_mode}",
    "defaultRole":    "${user_config.default_role}",
    "credentials":    { "username": "${user_config.auth_username}", "password": "${user_config.auth_password}" },
    "credentialsByRole": { ...roles object from auth-resolve.mjs, or omit if not per-role... },
    "storageStatePath": "${user_config.auth_storage_state_path}"
  }
}

The runner writes <runDir>/result.md and embeds every screenshot inline.

Phase 3 — Append to the run history + render dashboard

Run, in order:

node "${CLAUDE_PLUGIN_ROOT}/scripts/results-index.mjs" append "QA-tests/results/runs/<runId>"
node "${CLAUDE_PLUGIN_ROOT}/scripts/render-report.mjs"   "QA-tests/results/runs/<runId>"

The first updates QA-tests/results/history.json, QA-tests/results/latest.json, and QA-tests/results/by-task/<taskId>/latest.json. The second writes QA-tests/results/runs/<runId>/report.html — a self-contained HTML dashboard for the run (single-task runs render with auto-refresh off since the work is already done by the time we get here).

Phase 4 — Print summary

Print the runner's single-line JSON summary plus the absolute paths to both result.md and report.html.

Result schema (enforced by the runner)

See agents/qa-test-runner.md for the full result.md template. Highlights:

• | Result | PASS / FAIL | machine-parseable header table.
• One ### TC-NN: section per test case from the task file.
• Screenshots embedded as ![alt](TCNN-step.png) — never a table.
• Trailing ## Defects Found section using DEF-<slug> ids.