Stats
Actions
Available In
Tags
Plugin
inspect-skills
Skills for working with Inspect AI and its ecosystem packages (Evals, Flow, Scout, Viz, SWE, Harbor, sandboxes). Covers ecosystem routing, eval log reading, and live eval babysitting via inspect ctl.
What's Inside
Skills3
babysitting-evals
/babysitting-evals
Use for monitoring and diagnosing RUNNING Inspect AI evaluations via the `inspect ctl` command-line control channel. Triggers when: watching a live eval for stalls/errors/retry failures, checking progress of an active eval process, launching an eval you'll monitor in parallel, or finding problematic samples in a running eval. Does NOT apply to: writing new eval tasks or scoring logic, refactoring existing eval tasks, fixing code bugs unrelated to live eval monitoring, or analyzing completed eval logs.
map-inspect-packages
/map-inspect-packages
Use whenever the user is working with Inspect AI or any of its ecosystem packages (Evals, Flow, Scout, Viz, SWE, Harbor, Sandboxes). Maps concerns to packages and points at each package's docs before calling its API.
reading-logs
/reading-logs
Use whenever the user needs to read, inspect, or process Inspect AI eval log files (`.eval` or `.json`). Covers the Python API (`read_eval_log`, `read_eval_log_samples`, `read_eval_log_sample_summaries`, `list_eval_logs`), `header_only` mode, the critical anti-pattern of unzipping `.eval` files, and memory-safe patterns for large logs.
README
Inspect Skills
Coding agent skills for Inspect AI and its ecosystem (inspect_ai, inspect_evals, inspect_flow, inspect_scout, inspect_viz, inspect_swe, inspect_harbor, sandboxes).
Install
Claude Code (recommended)
/plugin marketplace add meridianlabs-ai/inspect-skills
/plugin install inspect-skills@meridian
Auto-update (recommended)
Third-party marketplaces are manual-update by default. To get our changes automatically (every push to main lands in your next Claude Code session), toggle auto-update on for the meridian marketplace from /plugin > Marketplaces, or add to your ~/.claude/settings.json:
{
"extraKnownMarketplaces": {
"meridian": { "autoUpdate": true }
}
}
To update manually instead: /plugin marketplace update meridian then /reload-plugins.
Other agents (Codex, Cursor, GitHub Copilot, etc.)
npx skills add meridianlabs-ai/inspect-skills --agent <agent> -g -y
-y installs every skill non-interactively; add --skill <skill-name> to install just one. Updates: npx skills update. Run npx skills add --help for the full list of supported agents.
Skills
| Skill | When it fires |
|---|---|
map-inspect-packages | Starting work in the Inspect AI ecosystem. Picks the right package for a given task and points at its docs. |
reading-logs | Reading, inspecting, or processing Inspect AI eval log files (.eval or .json). Covers the inspect_ai.log API and memory-safe patterns. |
babysitting-evals | Monitoring and diagnosing running Inspect AI evals via inspect ctl (the read-only control-channel CLI). Covers stall diagnosis, error triage, launch-and-watch workflows, and graceful stops. |
Stats
Version1.0.0
Stars0
MaintenanceExcellent
LicenseMIT
Last Commit
Added
Available In
Tags
Categories
