Skip to content

[bug] AI-generated comment 中的 @mention 可能繞過 tagging-collaborators 5-step protocol (sister concern from #116) #117

Description

@kiki830621

Problem

#116 sister concern 浮出: IDD skills 在 AI-generated comment / diagnosis body 中 cross-reference 其他人(@user mention)時,可能繞過 rules/tagging-collaborators.md 5-step 協定。

範例風險場景:

  • idd-verify 6 個 agent 各自在 raw output 寫 @codex / @claude-reviewer 內部 reference,被 GitHub 解析成 user mention → 通知無辜的同名 user
  • Sister concern surfacing audit 提及 @assignee 等動態值
  • Linked-context sister sweep 引用原 conversation 中含 @xxx token

rules/tagging-collaborators.md 5-step 主要約束 human-initiated mention,沒明確涵蓋 AI 內部互相 reference 的 case。

Type

bug(latent — 未確認觸發,但 audit 發現 path 存在)

Expected

AI-generated content 在 post 到 GitHub 前,所有 @\w+ token 走 rules/tagging-collaborators.md 5-step verification,或被 escape 成 `@xxx` literal(若是內部 codename 而非真實 user)。

Actual

未 audit。猜測各 skill 各自處理,缺 cross-skill 統一 enforcement。

Source

surfaced during /idd-diagnose #116 sister concern surfacing (Step 3.6)


Current Status

Phase: closed
Last updated: 2026-07-05 by /idd-verify (via /idd-all batch drain)

Key Decisions

Metadata

Metadata

Assignees

No one assigned

    Labels

    bugSomething isn't working

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions