Full feature-development pipeline for this repo — isolated worktree, spec, implementation plan, TDD build, quality gates, then a review→fix loop that only stops when review passes. Each stage delegates to a named project subagent (spec-researcher, pattern-scout, pipeline-implementer, pipeline-simplifier, pipeline-reviewer) with a stage-appropriate model. Use when starting any new MCP tool, feature, or multi-file change from scratch; when the user says "new tool/feature" (any language), "run the pipeline", or asks to develop something "the same way as list_test_records". Triggers on `/dev-pipeline`. NOT for one-file tweaks, doc edits, or work already mid-flight — jump to the matching stage instead.
Deploy skill for mcp-server-polarion. Automate version bump, tag create, PyPI publish. Triggers on `/deploy` or any release/version bump/tag request.
Shrink LLM-facing MCP tool descriptions — `@mcp.tool` docstrings and `Field(description=...)` — to eval-failure boundary via compress→eval→compress loop with snapshot+rollback, all gates green. Use when asked shrink/compress/optimize/token-diet tool descriptions, docstrings, or param descriptions. NOT for ordinary code comments (use compress-code-comments). Triggers on `/shrink-mcp-tool-docs`.
Compress verbose developer code comments and dev docstrings in .py files, Google-style — keep the why, drop the what, one line per point; gated on a comment-only AST diff. Use when asked to tidy/shrink/clean up code comments, cut comment verbosity, or trim over-explained docstrings. NOT for `@mcp.tool` docstrings or `Field(description=...)` (LLM-facing — use shrink-mcp-tool-docs). Triggers on `/compress-code-comments`.