Tool-calling agent with guardrails

• Map every rubric criterion to a file, test, or artifact.
• Include reproducible install, run, and verification commands.
• Record one meaningful decision, alternatives, and tradeoffs.
• Show representative output and state known limitations.

# Tool-calling agent with guardrails

## Goal
An agent that uses 3+ tools (search, calculator, file ops) with schema-validated arguments, retries, a hard step budget, and an audit log of every tool call. Demonstrate one failure case handled gracefully.

## Acceptance criteria
- [ ] Tool contract safety: Validated args; no arbitrary code execution paths.
- [ ] Loop control: Step budget + termination proofs; no runaway loops.
- [ ] Observability: Audit log readable enough to debug a bad run.
- [ ] Failure handling: A real failure case caught and recovered.

## Verification
- [ ] Document installation and run commands.
- [ ] Record automated checks and their exact commands.
- [ ] Add representative output, screenshots, or a short demo where useful.
- [ ] Confirm failure paths and known constraints.

## Decision log
### Decision title
- Context:
- Choice:
- Alternatives considered:
- Tradeoffs:

## Evidence
- Link each rubric criterion to the file, test, or artifact that demonstrates it.

## Known limitations
- List what is intentionally out of scope and what should be improved next.