pgagarinov

Stop the running supervisor and all researcher variants

Live tree view of supervisor via cron — agent collects stats every 30s

Dashboard: supervisor status, researcher variants, and metric history

Remove all supervisor leftovers: clones, logs, state, backups

Start the supervisor as a background process

Delete all deployed SAR repos and clean temp files

Reset the target to its initial seed state

Run REAL end-to-end integration tests for the SAR system

Audit codebase for consistency issues, anti-patterns, and missing test coverage

Python API coding conventions for this BookStore project. Use when writing or reviewing Python code.

Book publishing domain knowledge including ISBN, BISAC codes, and industry pricing standards.

Generate pytest test suite for a given module

Read-only mode for exploring code safely. Restricts Claude to read, search, and glob operations only.

Run a security-focused code review identifying vulnerabilities

Explain code using visual ASCII diagrams and analogies

Summarize changes in the current branch or working directory

Start the research loop and begin monitoring

Clean supervisor state, logs, and temp files

Read, edit, and diff .claude prompt assets (SKILL.md, agent definitions) in the supervised repo.

Stop the research loop and capture final snapshot

Clean research loop state and results

Fix wrong-chunk-index selection in cross-doc diversity

Read, edit, and diff .claude prompt assets (skills, agents, rules) in the target repo.

Build or load cached USearch index from QASPER, run eval pipeline, report metrics

Convert documents (PDF, Word, PowerPoint, Excel) in the vault to Markdown. Use when the user asks to convert documents, make files searchable, or index non-markdown files.