feat(agentv-bench): add executor subagent for non-CLI targets#808

Merged

christso merged 15 commits intomainfrom

feat/797-executor-subagent-skill

Mar 28, 2026

Commits on Mar 28, 2026

feat(agentv-bench): add executor subagent for non-CLI targets (#797)

christso
and
claude
committed
fix(agentv-bench): remove tools restriction from executor subagent

christso
and
claude
committed
docs(agentv-bench): add metadata field to executor input.json docs

christso
and
claude
committed
refactor(pipeline): align input.json fields with eval YAML and SDK

christso
and
claude
committed
refactor(pipeline): rename questionText to inputText for consistency

christso
and
claude
committed
refactor(pipeline): remove legacy question/reference_answer from grader payload

christso
and
claude
committed
fix(agentv-bench): update stale input_text/input_messages references

christso
and
claude
committed
fix(agentv-bench): simplify executor — remove workspace-dir parameter

christso
and
claude
committed
docs(agentv-bench): add workspace detection guidance for executor subagents

christso
and
claude
committed
docs(agentv-bench): clarify workspace detection — only matters for workspace-dependent evals

christso
and
claude
committed
fix(pipeline): extractInputText should use all messages, not just first

christso
and
claude
committed
docs(eval): fix inputText jsdoc — includes all messages, not just first

christso
and
claude
committed
docs(eval): fix expectedOutputText jsdoc — last message content, not all

christso
and
claude
committed
fix(pipeline): use @[role] pattern for multi-message inputText

christso
and
claude
committed
fix(pipeline): align @[role]: format with artifact-writer convention

christso
and
claude
committed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat(agentv-bench): add executor subagent for non-CLI targets#808

feat(agentv-bench): add executor subagent for non-CLI targets#808
christso merged 15 commits intomainfrom
feat/797-executor-subagent-skill

Commits on Mar 28, 2026

feat(agentv-bench): add executor subagent for non-CLI targets (#797)

fix(agentv-bench): remove tools restriction from executor subagent

docs(agentv-bench): add metadata field to executor input.json docs

refactor(pipeline): align input.json fields with eval YAML and SDK

refactor(pipeline): rename questionText to inputText for consistency

refactor(pipeline): remove legacy question/reference_answer from grader payload

fix(agentv-bench): update stale input_text/input_messages references

fix(agentv-bench): simplify executor — remove workspace-dir parameter

docs(agentv-bench): add workspace detection guidance for executor subagents

docs(agentv-bench): clarify workspace detection — only matters for workspace-dependent evals

fix(pipeline): extractInputText should use all messages, not just first

docs(eval): fix inputText jsdoc — includes all messages, not just first

docs(eval): fix expectedOutputText jsdoc — last message content, not all

fix(pipeline): use @[role] pattern for multi-message inputText

fix(pipeline): align @[role]: format with artifact-writer convention