Hi, I maintain SACP, a small text-first receipt layer for AI agent work:
https://github.com/aDragon0707/sacp
I noticed this project focuses on agent observability and replay tooling. That seems like a good fit for a docs-only example showing how an agent's final completion claim can be represented as a receipt.
SACP does not replace replay, traces, or observability systems. It documents the final work state:
- what the agent claimed
- what evidence supports it
- who owns the next step
- whether human approval is required
Would you accept a tiny docs-only PR adding a SACP-style receipt example for an agent completion claim?
No runtime change, no dependency, no adoption requirement. It would just be an example file such as:
docs/sacp_receipt_example.md
The concrete failure mode is:
an agent says "done" or "tests passed", but there is no explicit receipt tying the claim to evidence, next owner, and approval boundary.
Hi, I maintain SACP, a small text-first receipt layer for AI agent work:
https://github.com/aDragon0707/sacp
I noticed this project focuses on agent observability and replay tooling. That seems like a good fit for a docs-only example showing how an agent's final completion claim can be represented as a receipt.
SACP does not replace replay, traces, or observability systems. It documents the final work state:
Would you accept a tiny docs-only PR adding a SACP-style receipt example for an agent completion claim?
No runtime change, no dependency, no adoption requirement. It would just be an example file such as:
The concrete failure mode is:
an agent says "done" or "tests passed", but there is no explicit receipt tying the claim to evidence, next owner, and approval boundary.