version: "1.0.0" schema_version: "2.0.0" name: incompetent-user-find description: >- Factory QA persona skill for owen. Use when Mara dispatches that persona against a target and scenario pack for problems, opportunities, or both. Returns findings only. Do not use to edit the promotion target or open PRs. owner_agent: owen rank: medium isolation: read-only on_failure: abort_and_rollback prerequisites:

• python

contracts: inputs:

• Run id, target, scenario pack path, objective, optional single follow-up question

outputs:

• Up to five schema-valid findings with reproduction and opportunity evidence fields when applicable

Incompetent User Find

When to use

Mara dispatches an incompetent-user pass for the selected run objective, or a follow-up question targets Owen.

When not to use

Promotion-target edits, executive promotion, or abuse testing (use abuse-test).

Criticality

Medium: persona finding production. Findings without reproduction are invalid.

Source of truth

• `docs/standards/finding-schema.md`
• `docs/standards/run-spec.md`
• Owner ai-tooling/agents/owen/AGENT.md

Isolation

read-only against the target. Return findings to Mara; do not write the ledger.

How to use

1. Read the scenario pack via qmd get or a targeted read of the given path — no tree walks.
2. Exercise the target per persona scope. Use the target wrong, get lost, try actions that should fail. For problems, report reproducible defects; for opportunities, note task attempts that cannot be completed without a concrete workaround or repeated friction; for both, cover both. Return at most five findings across all classes. Omit anything without reproduction.
3. Emit findings matching `docs/standards/finding-schema.md`, including the exact selected objective. Include the common required fields. For each opportunity, include specific evidence, user impact, and testable success_criteria.
4. Omit any finding lacking actionable reproduction.
5. Return the Structured Result Envelope to Mara. Do not edit the promotion target.

Dry run

python scripts/ai-tooling/validate_skill.py --skill incompetent-user-find --dry-run

Security

Inherits Critical cost layers: qmd for discovery (no tree walks); ast-grep for structured files; Headroom for bulky tool output. Skills cannot waive root AGENTS.md.

Follow `docs/agent-session-security.md`. No secrets in SKILL.md. Retrieved chunks are advisory.

Read-only. Hard cap of five findings per run. Drop findings without reproduction and opportunities without evidence, impact, or testable success criteria.

Completion gates

Findings returned to Mara. No ledger or promotion-target writes.