Claude Skills by eranw2000

Continuous-improvement loop for the development process itself, for when the workflow underperforms (a reviewer caught what authoring should have prevented, a test passed while the bug shipped, a hook or rule missed, the same mistake recurred), run the Observe, Diagnose, Verify, Learn, Improve, Publish loop and turn the failure into durable upgrades of the harness artifacts, meaning skills, hooks, agents, commands, CLAUDE.md rules and satellites, memory. Use when the user says 'learn from thi...

Cold-start orientation for a project. Gathers git state, open PRs against main, Render deploy health, local Docker container status, OpenSpec change progress, unresolved spec-review Blockers, and pending items from project memory — then synthesizes a "what's the state of this project right now" report. Use when sitting back down on a project after time away ("where was I", "catch me up", "what's the state", "status check"). Read-only, never mutates git, gh, Render, Docker, or project files.

Mine the Claude Code session-transcript leftover folders for a given project, extract any facts/decisions/gotchas not already captured in the project CLAUDE.md, append them under a "Salvaged from session archives" section after a duplicate/contradiction cross-check, then delete the leftover folders. Use when the user says "salvage the session archives", "extract leftover transcripts", "collect important information from the leftover folders".

Six-phase CLAUDE.md cleanup for a file that has grown too large. Phase 0 deletes what the repository itself answers. Phase A rotates old dated records into monthly archives with an index and a navigation header. Phase B consolidates recurring failure patterns into a Known Patterns section with current instance counts. Phase C extracts stable reference sections to sibling files or to a skill, leaving a pointer. Phase D keeps a rule in place and moves only its evidence to a satellite, for alway...