agent-execution-receipts
Inspect signed Codex execution receipts, import provider audit events, and explain whether an action has local evidence, a candidate correlation, or a provider binding.
来源信息
- 仓库
- writer/cerebro
- 最近来源活动
- 2026年7月16日 08:39
- 检测到的 SKILL.md 语言
- 英语
- 星标
- 14
- 分支
- 2
安装方式
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
检查来源文件
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
正在显示 SKILL.md
SKILL.md
来源说明 · 只读预览- name
- agent-execution-receipts
- description
- Inspect signed Codex execution receipts, import provider audit events, and explain whether an action has local evidence, a candidate correlation, or a provider binding.
# Agent Execution Receipts
Use this skill when the user asks who or what ran a command, whether an agent action reached a provider, or which agent-originated changes lack provider attribution.
## Procedure
1. Open the Cerebro Agent Receipts app or read `receipts.ndjson` from the configured receipt directory.
2. Verify each receipt signature and the append-only digest chain before using it as evidence.
3. Import a CloudTrail `LookupEvents` response when provider observations are available.
4. Report one of three states without collapsing them:
- `Local evidence only`: a signed local receipt exists.
- `Candidate correlation`: one provider event has a one-to-one evidence match, but no trusted binding.
- `Provider bound`: an authenticated provider event passes the configured account, dedicated role, action ID, action, and time checks.
5. Start from the provider-event population and identify every provider mutation without a completed one-to-one action match.
Never treat a process name, user agent, `startedBy` value, local transcript, session-wide identifier, or user-imported JSON as provider-bound attribution.
在 GitHub 查看