Audit receipt
Use when an agent claims a review, edit, or run is complete and you need machine-readable proof of scope, checks, and gaps.
PR reviewagent editsCI evidenceReceipt gallery for agent audit work
Pick a receipt shape by the risk you need to prove, not by the tool name. Each template records what was checked, what stayed unverified, and whether the run should be trusted.
Templates
Profile exposure receipt
Use when a workflow depends on tools that may be hidden by restricted profiles, disabled environments, or missing permissions.
tool profilesblocked toolssetup driftLarge audit report receipt
Use for multi-agent or long audit reports where the summary can look complete while appendices, verifier counts, or drift risk remain unresolved.
multi-agent auditappendicesdrift riskMCP team audit receipt
Use when a team shares MCP servers or tool connectors and needs to capture identity, allowlists, redaction state, and offboarding gaps.
MCP teamstool callsaccess reviewInstruction diagnosis receipt
Use when trace-backed feedback says a user instruction was ambiguous, underspecified, or missing evidence requirements.
trace-backed feedbackinstruction qualityagent memoryWhich one should I use?
| Problem | Best starting receipt | Paid-pilot deliverable |
|---|---|---|
| An agent says a code review is done, but you cannot tell what was actually inspected. | Audit receipt | A reusable review workflow with pass/fail/incomplete criteria and a sample completed receipt. |
| A workflow fails only on some machines because a tool or profile is hidden. | Profile exposure receipt | A setup checklist plus receipt fields that expose required tools, visible tools, and blocked paths. |
| A large generated audit report looks impressive but may hide missing appendices or unverified claims. | Large audit report receipt | A report contract with verifier counts, appendix publication state, and drift-risk handling. |
| A team shares MCP or tool-call infrastructure and needs access/accountability records. | MCP team audit receipt | A team-specific receipt schema for identity, allowlist, redaction, and offboarding evidence. |
| A trace-backed analyzer says the user's implementation instruction caused ambiguity or rework. | Instruction diagnosis receipt | A diagnosis contract with evidence refs, quote state, negative control, extractor source, and supported/weak/incomplete verdict. |