`WorkspaceWithBoard.save_findings` posts to blackboard with hardcoded solver_id and double-writes

[06mia]

WorkspaceWithBoard.save_findings posts to blackboard with hardcoded solver_id and double-writes

Severity: Medium
File: src/cain_agent/workspace_ext.py:55-63

save_findings overrides the base class to sync findings into the blackboard, but it hardcodes solver_id="legacy" for every finding regardless of origin, and calls post_finding which appends to self.blackboard._memory.findings:

def save_findings(self, findings: list[object]) -> None:
    super().save_findings(findings)
    for f in findings:
        finding = Finding(**f) if isinstance(f, dict) else f
        self.blackboard.post_finding(finding, solver_id="legacy")

Every call to save_findings (which happens in both the test handler and the validation pipeline) re-posts all findings to the blackboard, causing unbounded duplication in self._memory.findings. Additionally, the Finding dataclass from multi_agent/types.py has different fields than the pipeline Finding from findings.py, so Finding(**f) will raise TypeError for any dict that was serialized from the pipeline model (e.g., missing evidence_hash, extra result field).

Why it matters

Running the pipeline through WorkspaceWithBoard causes TypeError crashes on the first save_findings call after findings exist, and even if the schemas matched, the blackboard findings list would grow quadratically with each pipeline re-run.

[cdxiaodong]

确认属实,三点全部处理(main 665e3b9,合并于 4096e00):① 双写 — save_findings 现按 finding_id 与黑板现有 findings 查重,重复 run 不再膨胀;② 类型不匹配 — dict → board Finding 改走字段白名单映射(_to_board_finding),pipeline 侧多余字段(如 result)忽略、坏条目跳过,TypeError 不再外溢;③ solver_id 不再硬编码 legacy,取 finding 自带值缺省回落 workspace。Issue 暂保持开放。