WIP: stage all uncommitted work — sprints 46-221, graduation headers, specialist fleet, test steps 909-1988
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
This commit is contained in:
41
datasets/run_specs/runspec_wrong_tool_selection_04.md
Normal file
41
datasets/run_specs/runspec_wrong_tool_selection_04.md
Normal file
@@ -0,0 +1,41 @@
|
||||
# Run Spec: wrong_tool_selection_04
|
||||
|
||||
## Context
|
||||
|
||||
Synthetic MCP data-collection case.
|
||||
family = wrong_tool_selection
|
||||
variant = 4
|
||||
|
||||
## Objective
|
||||
|
||||
Exercise wrong_tool_selection behavior and produce deterministic stage-labeled traces.
|
||||
|
||||
## Anti-Pattern To Detect
|
||||
|
||||
- Select semantically adjacent tools for the wrong workflow phase.
|
||||
|
||||
## Target Behavior
|
||||
|
||||
- Use phase-correct ordering: intake -> generate -> queue -> validate.
|
||||
- Keep tool-call chain bounded and auditable.
|
||||
- Preserve deterministic output ordering.
|
||||
|
||||
## Requirements
|
||||
|
||||
- Workspace-relative paths only.
|
||||
- Deterministic tool sequencing.
|
||||
- Valid JSON arguments for each tool invocation.
|
||||
|
||||
## Acceptance Criteria
|
||||
|
||||
- At least one tool_selection event.
|
||||
- At least one tool_execution event.
|
||||
- No malformed MCP payloads.
|
||||
- No unbounded loops.
|
||||
|
||||
## Data Capture Tags
|
||||
|
||||
- family: wrong_tool_selection
|
||||
- variant: 4
|
||||
- capture_mode: run_spec
|
||||
- expected_failure_signature: wrong_tool_selection
|
||||
Reference in New Issue
Block a user