Use when writing or reviewing ad-hoc research SQL in a dbt-ClickHouse project, or running federated queries via postgresql() from ClickHouse for data-quality investigations — covers CTE-wrapping of federation, dedup via limit 1 by, flow analysis with groupUniqArray + arraySort, runtime-computed boundary dates, hybrid PG/CH source selection, source-of-truth citation, and style.
EnglishUse when you notice a divergence between actual session behavior and codified instructions (one of the 8 drift triggers) — captures it as an immutable drift-log entry under .claude/drift-log/open/ so good conventions can be promoted later.
EnglishUse when documenting a meeting from an SRT audio transcript — produces structured meeting notes with decisions, action items, and discussion…
EnglishUse when inspecting, probing, or reporting on any credential-bearing file (.mcp.json, .env, .pem, ~/.git-credentials, ~/.netrc, a secret store) or when clearing a "no secrets" / security-sweep verdict - the mechanics that keep a well-intentioned command from leaking a value through tool output, a diff echo, a query error, or a shallow history scan. The high-level policy (where secrets live, rotation) is assumed; this is the HOW.
EnglishUse for any research, validation, or fact-check task -- before answering a load-bearing question or committing to an approach. Orients via local sources and knowledge tools, fans out parallel subagents per source, cross-validates via MCP servers and web research, and cites sources.
EnglishUse when the user wants to review AI-agent instruction files (CLAUDE.md, skills, drift-log, hooks) for pollution, duplication, dead references, contradictions, or architectural debt. Manual on-demand only. Outputs a committed markdown report; does no auto-fix. Always run together with /claude-md-management:claude-md-improver for full coverage.
EnglishUse when triaging the drift-log — promoting open entries to applied, checking staleness, compacting applied entries, or turning an applied insight into a codified rule.
EnglishUse when a project CLAUDE.md is bloated, predates the canonical skeleton, or duplicates the shared home layers (the ai/core behaviour-protocol and the personal ~/.claude/CLAUDE.md) - thins it to the 10-section skeleton by removing content the home layers already provide while preserving every project-specific fact. The FIX companion to reviewing-agent-instructions (which only audits). Do NOT use it to change project rules or to edit a CLAUDE.md that is already thin and conformant.
EnglishUse when driving any web UI through the Playwright MCP -- logging into SSO-gated portals, filling Monaco or textarea editors, extracting page content, or operating internal tools via browser automation. Covers session/login handling, reliable text input, snapshot-vs-screenshot, and token-efficient extraction.
English
AGENT SKILLS ATLASIncludes Clickhouse Query Best Practices, Creating Drift Logs, Documenting Meetings, and other workflows.
Claude Code
Codex
Cursor
GitHub Copilot
Gemini CLISupported by the Atlas install prompt. Skill-specific requirements vary.
Pack stats
What’s included
Clickhouse Query Best Practices
Use when writing or reviewing ad-hoc research SQL in a dbt-ClickHouse project, or running federated queries via postgresql() from ClickHouse for data-quality investigations — covers CTE-wrapping of federation, dedup via limit 1 by, flow analysis with groupUniqArray + arraySort, runtime-computed boundary dates, hybrid PG/CH source selection, source-of-truth citation, and style.
Creating Drift Logs
Use when you notice a divergence between actual session behavior and codified instructions (one of the 8 drift triggers) — captures it as an immutable drift-log entry under .claude/drift-log/open/ so good conventions can be promoted later.
Documenting Meetings
Use when documenting a meeting from an SRT audio transcript — produces structured meeting notes with decisions, action items, and discussion summary
Handling Secrets Safely
Use when inspecting, probing, or reporting on any credential-bearing file (.mcp.json, .env, .pem, ~/.git-credentials, ~/.netrc, a secret store) or when clearing a "no secrets" / security-sweep verdict - the mechanics that keep a well-intentioned command from leaking a value through tool output, a diff echo, a query error, or a shallow history scan. The high-level policy (where secrets live, rotation) is assumed; this is the HOW.
Researching Rigorously
Use for any research, validation, or fact-check task -- before answering a load-bearing question or committing to an approach. Orients via local sources and knowledge tools, fans out parallel subagents per source, cross-validates via MCP servers and web research, and cites sources.
Reviewing Agent Instructions
Use when the user wants to review AI-agent instruction files (CLAUDE.md, skills, drift-log, hooks) for pollution, duplication, dead references, contradictions, or architectural debt. Manual on-demand only. Outputs a committed markdown report; does no auto-fix. Always run together with /claude-md-management:claude-md-improver for full coverage.
Reviewing Drift Logs
Use when triaging the drift-log — promoting open entries to applied, checking staleness, compacting applied entries, or turning an applied insight into a codified rule.
Thinning Agent Instructions
Use when a project CLAUDE.md is bloated, predates the canonical skeleton, or duplicates the shared home layers (the ai/core behaviour-protocol and the personal ~/.claude/CLAUDE.md) - thins it to the 10-section skeleton by removing content the home layers already provide while preserving every project-specific fact. The FIX companion to reviewing-agent-instructions (which only audits). Do NOT use it to change project rules or to edit a CLAUDE.md that is already thin and conformant.
Using Playwright
Use when driving any web UI through the Playwright MCP -- logging into SSO-gated portals, filling Monaco or textarea editors, extracting page content, or operating internal tools via browser automation. Covers session/login handling, reliable text input, snapshot-vs-screenshot, and token-efficient extraction.
Quick start
- Choose your skills. Select individual skills above, or choose the entire pack.
- Copy the install prompt. Use Copy install prompt above the list.
- Send it to your agent. Paste into your coding agent’s chat. Your agent reviews the source and guides the setup.
Try it with your agent
After installation, start with a request like:
Use the Clickhouse Query Best Practices skill to help with my task.
Ask me for the context you need before getting started.Requirements & license
- A coding agent that can access your project files and install skills.
- Your agent checks each skill’s instructions, dependencies, and permissions during setup.
- Reported license for the skills shown: MIT. The author’s terms apply.