Add delegation and pre-action authorization attributes for agent spans
维护者通常 1 天内回复
评估
- 难度
- 4/5
- 预计耗时
- 3-5 天
- 新手友好度
- 38/100
- Issue 类型
- 功能
- 描述清晰度
- 基本清楚
- 活跃度
- 活跃
- 技术栈
- python
调研方向
Locate the existing handoff span implementation and the attribute registry, then trace where pre-action governance telemetry is recorded. Compare the proposed fields with the project’s conventions for scheme-neutral span attributes and RFC 8693 or signed-delegation data. Done means the agreed attributes are represented and emitted without including credentials or policy text, with coverage confirming their values across delegation and authorization cases.
由索引模型根据 Issue 内容生成。
描述
Following the thread on pre-action governance telemetry: the handoff span records source and target agents, but not the authority the target is acting under. In multi-hop delegation that is the question auditors ask first: on whose behalf, and was this action within what was delegated?
A small, scheme-neutral set of attributes could cover it without putting credentials or policy text into traces, in line with the references-and-hashes approach discussed above:
- gen_ai.delegation.root.id: the principal at the root of the chain
- gen_ai.delegation.actor.id: the agent acting now
- gen_ai.delegation.depth: number of delegation hops
- gen_ai.delegation.chain.hash: hash of the delegation chain as verified, so an auditor can fetch and re-verify it out of band
- gen_ai.authz.decision: allow or deny, recorded before the tool call executes
- gen_ai.authz.reason: a short machine-readable code for a denial
These map directly onto RFC 8693 token exchange (sub and the nested act claims) and onto signed delegation formats. We maintain one such format, VAID (https://github.com/solara-associates/vaid), and a bridge from verified VAID chains to RFC 8693 JWTs, so we could emit these attributes in practice and report back on what works.
Happy to draft a PR against the attribute registry if this direction fits.
- 主要语言
- Python
- 星标
- 7.5k
- 派生
- 1.2k
- 平均合并
- 1 天 23 小时
- 30 天内合并 PR
- 3
环境准备
- 没有 Dockerfile 或 Docker Compose 文件
- 有 Pull Request 模板
- 阅读贡献指南
从这里开始
- 先读完整个 Issue,再读项目的贡献指南。
- 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
- Fork 仓库,在一个分支上完成修改。
- 提交 Pull Request,并在描述里引用这个 Issue 编号。
traceloop/openllmetry 的其他 Issue
-
🐛 Bug Report: OpenAI Responses: with_raw_response.create() raises when the body can't be parsed, though the uninstrumented call returns可能已有人在做 关联的 PR 仍在进行中或已合并。 未关闭bug python
难度 2/5 1-3 小时 新手友好度 76/100
traceloop/openllmetry#4717 ·
维护者通常 1 天内回复
-
🐛 Bug Report: Incomplete cache token migration after #4243 breaks Bedrock and Groq instrumentation可能已有人在做 @Quantum-Enigma-dot 于 21 天前认领。 未关闭
难度 2/5 1-3 小时 新手友好度 76/100
traceloop/openllmetry#4665 · 1 条评论 ·
维护者通常 1 天内回复
-
🐛 Bug Report: OpenAI instrumentation drops gen_ai.usage.reasoning_tokens on streamed chat completions可能已有人在做 @jayfeng7 于 23 天前认领。 未关闭
难度 2/5 1-3 小时 新手友好度 82/100
traceloop/openllmetry#4660 · 1 条评论 ·
维护者通常 1 天内回复
-
Bug Report: Failed dataset deletes still remove local rows and columns可能已有人在做 @inchang-ing 于 32 天前认领。 未关闭
难度 2/5 1-3 小时 新手友好度 88/100
traceloop/openllmetry#4630 ·
维护者通常 1 天内回复
-
🐛 Bug Report: LiteLLM instrumentation does not record seed, stop, frequency_penalty, presence_penalty可能已有人在做 @shantam77 于 38 天前认领。 未关闭
难度 2/5 1-3 小时 新手友好度 85/100
traceloop/openllmetry#4598 · 2 条评论 ·
维护者通常 1 天内回复
查看 traceloop/openllmetry 的全部 Issue
相似的 Issue
-
Layout accepts a step threshold of 0 or less, then in-toto-verify crashes with StopIteration可能已有人在做 @BahaaMattar 今天认领。 未关闭
难度 2/5 1-3 小时 新手友好度 76/100
-
难度 2/5 1-3 小时 新手友好度 65/100
维护者通常 1 天内回复
-
needs discussion
难度 2/5 1-3 小时 新手友好度 78/100
openfga/openfga.dev#1436 · 1 条评论 ·
维护者通常 2 天内回复
-
难度 2/5 1-3 小时 新手友好度 82/100
维护者通常 1 天内回复
-
usb-device-hid: MouseInterface reports bInterfaceProtocol 1 (keyboard) instead of 2可能已有人在做 @Duhan07 今天认领。 未关闭
难度 2/5 1-3 小时 新手友好度 82/100
micropython/micropython-lib#1176 ·