From agent-steer
Label session outcomes (success/failure/partial) with failure mode tags for ground truth measurement
How this skill is triggered — by the user, by Claude, or both
Slash command
/agent-steer:labelThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Annotate sessions with outcome labels and failure mode tags for ground truth analysis.
Annotate sessions with outcome labels and failure mode tags for ground truth analysis.
Label the current session:
!node "${CLAUDE_SKILL_DIR}/../../dist/label-query.js" set --outcome=success
Label with failure tags:
!node "${CLAUDE_SKILL_DIR}/../../dist/label-query.js" set --outcome=failure --tags=thrashing,wrong_approach
Label with notes:
!node "${CLAUDE_SKILL_DIR}/../../dist/label-query.js" set --outcome=partial --tags=missing_context --notes="got stuck on auth flow"
Label a specific session:
!node "${CLAUDE_SKILL_DIR}/../../dist/label-query.js" set --outcome=failure --session=abc123 --tags=regression
Valid outcomes: success, failure, partial
Valid tags: thrashing, wrong_approach, missing_context, test_blindness, regression, other
Show all labels:
!node "${CLAUDE_SKILL_DIR}/../../dist/label-query.js" query
Filter by outcome:
!node "${CLAUDE_SKILL_DIR}/../../dist/label-query.js" query --outcome=failure
Filter by tag:
!node "${CLAUDE_SKILL_DIR}/../../dist/label-query.js" query --tag=thrashing
Limit results:
!node "${CLAUDE_SKILL_DIR}/../../dist/label-query.js" query --outcome=failure --limit=50
Present the output to the user exactly as shown. Do not interpret or modify the values.
npx claudepluginhub sust4in/agentsteerGuides reception of code review feedback: verify before implementing, avoid performative agreement, push back with technical reasoning when needed.
Design banners for social media, ads, website heroes, and print with multiple art direction options and AI-generated visuals.