
Claude Skills by lamenting-hawthorn
github.com/lamenting-hawthornSkill chaining engine. Runs a sequence of skills with auto-continue between autonomous steps and human checkpoints at defined gates. Use when asked to "run the chain", "chain skills", "full workflow", or when a multi-step workflow is needed. Example: /chain debug-fix-ship
Proactively discover and learn codebase conventions, patterns, and architecture. Scans the project for naming, structure, testing, error handling, and component patterns, then saves them to memory and CLAUDE.md. Use when asked to "learn this codebase", "discover conventions", "analyze patterns", or when starting work in an unfamiliar project.
Escalation protocol for when the agent is unsure, stuck, or about to take a risky action. Defines escalation tiers and decision boundaries. This skill is primarily used internally by other M2* skills, not invoked directly. But can be used as /escalate to configure escalation boundaries.
View quality metrics and trends across sessions. Shows success rates, common corrections, skill usage, and improvement trends over time. Use when asked to "show quality", "how am I doing", "dashboard", "show metrics", or "quality report".
Quick mid-session reflection. Lighter than /self-learn, designed to be run during work to check if the approach is on track. Checks assumptions, reviews progress against goals, and surfaces potential issues early. Use when asked to "step back", "are we on track", "reflect", "check yourself", or when the task feels like it's going sideways.
Post-session reflection skill. Reviews the conversation for feedback, corrections, discovered conventions, and quality signals. Extracts learnings and persists them to memory for future sessions. Use when asked to "reflect", "what did you learn", "save learnings", "self-learn", or at the end of a productive session. Proactively suggest at the end of sessions where significant work was done.
Pre-completion self-review. Before finishing a task, reviews own work for correctness, completeness, code quality, and adherence to project conventions. Use when asked to "review your work", "check yourself", "self-review", or before shipping significant changes. Proactively suggest before committing or shipping non-trivial changes.