santos

Code review, cleanup, comment deslopping, refactoring, debugging discipline, browser testing agent, and gate-keeper evaluation agent. PEP workflow moved to its own marketplace at the pep repo.

8 skills · 1 command · 2 agents

Persistent session memory: auto-checkpoints sessions (PreCompact/SessionEnd), /note-session writes daily logs, /rollup-day distills them, /recall searches the curated layer. No external deps.

1 skill · 3 commands · hooks

Git commit workflow commands: commit, push and PR, branch cleanup.

4 commands

PEP workflow: markdown-driven planning, investigation, and execution. Wave shapes, adversarial review, evidence-required gates. Folds in the spike (investigation) flow. Audit hook records what actually ran. Includes the grill skill (alignment interview + domain glossary) as the planning front door.

2 skills · 1 agent · hooks

Background skill-library maintenance. Weekly idle-gated pass that consolidates near-duplicate skills into umbrella skills with references/templates/scripts subfiles. Pre-run snapshots, rollback, pin. Ported from Hermes.

3 commands

Background review hook that learns from each session. Interval-gated per-turn reviews with frustration/save-this overrides, ported from Hermes agent's self-improvement loop.

1 skill · hooks

Multi-model AI consultation using cursor-agent. Supports consultation, research, and deliberation modes.

1 skill

Syncs Claude Code plugins to Cursor skills and commands directories with launchd watcher.

1 skill

Turn spoken or typed descriptions into editable .excalidraw diagrams, with canvas edits synced back into the graph.

1 skill

Search and summarise Claude Code conversation history. Find past discussions and generate work summaries.

1 skill

Logo design generation pipeline using Gemini image generation. Generates prompt variations, calls API, evaluates results, iterates, and produces brand kits.

1 skill

QA spine. qa-scenarios generates user-impact test scenarios from a code change, reading a layered repo-local QA.md failure-mode corpus discovered at runtime. Product-agnostic; grows the corpus as the user reacts to scenarios.

1 skill

Utilities for creating and validating Claude Code skills. Scaffolds new skills for plugins or repositories.

1 skill