test(cli): prove the assert_tool_call argument matchers can fail

[AminChirazi]

Milestone 2, criterion 6. Closes #247. Based on main.

The gap

The tool-NAME layer already has a red path — a_failing_assertion_refuses_the_trace (agent_flow_e2e.rs:778) demands a tool the agent never calls, and record refuses the trace. Nothing proved the where <path> <matcher> clauses can fail at all.

That is the layer carrying the most weight and the least evidence. docs/agent-testing.md calls argument assertions "usually where the bugs are", and names chained arguments — threading one tool's result into the next call — as the behaviour multi-step agents actually get wrong. Yet every where clause in the suite asserts an argument the model was always going to produce, so each passes whether the matcher works or not.

One guilty call, both halves of the vocabulary

The fixture is the right tool with the wrong city, and it violates two assertions at once:

• where city equals Nairobi — a value matcher; the value is Mombasa
• where city is absent — a presence matcher; the key is there

Both refuse the record, and the helper asserts neither minted a trace. The presence case is worth its own test: a presence matcher that could not fail would let a flow assert the absence of anything at all and stay green.

Result

Both matchers behaved. No false green.

Verified locally

fmt --check = 0, clippy --all-targets -D warnings = 0, both tests = 0 — each captured directly, never through a pipe.

🤖 Generated with Claude Code

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.