Update to Riqor

{
  "description": "Use when an AI coding agent needs a repository-scoped Riqor run, ordered evidence trace, verification gate, or defensible completion claim.",
  "included_files": [],
  "name": "riqor-evidence",
  "skill_md_contents": "---\nname: riqor-evidence\ndescription: Use when an AI coding agent needs a repository-scoped Riqor run, ordered evidence trace, verification gate, or defensible completion claim.\n---\n\n# Riqor Evidence\n\nCompletion claims must follow observable verification from the same repository state.\n\n## Start a run\n\nUse one active run per repository and state the goal explicitly\n\n```bash\nriqor run start --goal \"<specific outcome>\" --path evidence-loop --profile assured\n```\n\nInspect state with `riqor run status --json`\n\n## During implementation\n\n- A successful recognized mutation creates pending verification evidence\n- A failed mutation does not create fresh success evidence\n- A later mutation invalidates earlier verification for completion purposes\n- Keep verification focused on the changed behavior first, then run the broader gate required by the repository\n- Do not represent prose, masked commands, or an agent statement as test evidence\n\n## Trace review\n\nUse `riqor trace show <run-id> --json` to inspect ordered events and `riqor trace export <run-id> --format jsonl` when machine-readable trace data is needed\n\nRiqor trace state is intentionally bounded and content-free. Do not add raw commands, prompts, command output, source text, or credentials to trace metadata\n\n## Completion\n\nRun the required verification after the final mutation, confirm the run is no longer verification-pending, then use `riqor run complete --json`\n\nIf verification fails, preserve the failure as evidence and continue the fix cycle. Never clear pending evidence by assertion alone\n"
}