Skip to content

fix(crab-pf): require test prompts to match provider input format#33

Merged
MrFlounder merged 1 commit intomainfrom
feat/crab-pf-verify-session
Feb 22, 2026
Merged

fix(crab-pf): require test prompts to match provider input format#33
MrFlounder merged 1 commit intomainfrom
feat/crab-pf-verify-session

Conversation

@MrFlounder
Copy link
Contributor

Summary

  • Add system prompt rule requiring test prompts in promptfooconfig.yaml to match the provider's expected input format
  • Fixes false-positive verifications where providers expecting structured input (JSON objects, keywords) pass smoke tests but fail eval with generic prompts like "Hello"

Test plan

  • Run crab pf against app 09 (expects JSON-structured prompts) — verify generated test cases are valid JSON
  • Run crab pf against app 11 (expects keyword prompts) — verify generated test cases use keywords like "summary"/"transactions"

🤖 Generated with Claude Code

Providers that expect structured input (JSON objects, keywords) fail
when given generic test prompts like "Hello, how are you?". Add a rule
to the system prompt so the agent generates test cases that match the
provider's expected input format.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
@MrFlounder MrFlounder force-pushed the feat/crab-pf-verify-session branch from f430ec1 to af0c812 Compare February 22, 2026 06:06
@MrFlounder MrFlounder merged commit b5017cb into main Feb 22, 2026
3 checks passed
@MrFlounder MrFlounder deleted the feat/crab-pf-verify-session branch February 22, 2026 06:07
This was referenced Feb 22, 2026
This was referenced Mar 18, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant