Update to XMemo

{
  "description": "Review a proposed plan against saved XMemo goals, decisions, constraints, and current evidence. Use for a plan critique or acceptance recommendation, not implementation progress verification.",
  "included_files": [
    {
      "relative_path": "agents/openai.yaml",
      "size_in_bytes": 406
    }
  ],
  "name": "xmemo-review-plan",
  "skill_md_contents": "---\nname: xmemo-review-plan\ndescription: Review a proposed plan against saved XMemo goals, decisions, constraints, and current evidence. Use for a plan critique or acceptance recommendation, not implementation progress verification.\n---\n\n# Review a Plan with XMemo\n\nTest a proposal against durable context before execution. Follow\n**scope → baseline → decompose → challenge → verdict → gate**.\n\n## 1. Define the review contract\n\nIdentify the plan version, outcome, project or scope, decision owner, risk\ntolerance, and required depth. Distinguish draft, proposed, approved, and\nimplemented states.\n\n## 2. Recover the authoritative context\n\nUse `recall_context` for the bounded decision set, `search_memory` for exact\nconstraints, `get_project_summary` for current project state, and `todo` for\nactive commitments. Retrieve only context that can affect the review.\n\nResolve conflicts by current user instruction, verified evidence, recency, and\nscope. Use `explain_memory` when a decisive match or its provenance is unclear.\n\n## 3. Decompose and challenge the plan\n\nMap goals, assumptions, dependencies, sequencing, ownership, rollback,\nverification, acceptance criteria, and unresolved choices. Check:\n\n- alignment with saved goals and constraints;\n- contradictions with settled decisions;\n- missing evidence, dependencies, migration, or recovery paths;\n- security, privacy, data-loss, compatibility, and operational risks;\n- testability, observability, and release gates;\n- duplicated completed work or ignored lessons;\n- cross-agent ownership and handoff boundaries.\n\n## 4. Return a formal review\n\nUse:\n\n1. **Plan and baseline reviewed**\n2. **Verdict** — `recommended`, `changes_requested`, `blocked`, or `rejected`\n3. **Critical findings**\n4. **Required changes**\n5. **Risks and assumptions**\n6. **Missing evidence or decisions**\n7. **Acceptance gate** — exact evidence required before approval or execution\n\nSeparate blocking findings from optional improvements. A favorable review is\na recommendation, not user approval or authorization to execute.\n\n## 5. Preserve the outcome carefully\n\nKeep the verdict and user approval distinct. If saving the review is authorized,\nsave its durable rationale with `remember` or `update_memory`. Use `todo`,\n`update_state`, or `record_event` only for the separately authorized outcomes.\nApproval of the plan alone does not authorize any of these writes.\n\nClarify unresolved write scope and return a receipt for authorized writes. Never\nclaim implementation or deployment merely because a plan passed review. Read\n`../xmemo-memory-steward/references/review-playbooks.md`,\n`../xmemo-memory-steward/references/memory-policy.md`, and\n`../xmemo-memory-steward/references/tool-routing.md` for detailed rules.\n\n## Shared boundaries\n\nBefore any tool call, read and apply\n[plugin boundaries](../xmemo-memory-steward/references/plugin-boundaries.md).\nUse [tool routing](../xmemo-memory-steward/references/tool-routing.md) for unfamiliar\noperations, [temporal queries](../xmemo-memory-steward/references/temporal-queries.md)\nfor dates or chronology, and [memory policy](../xmemo-memory-steward/references/memory-policy.md)\nbefore writes. Reuse existing authorization; clarify only unresolved scope.\nReport failed or unconfirmed operations without claiming success.\n"
}