Agent skills

Let your agents inspect, search, and evaluate their own history.

thirdeye ships five Claude-Code-format skills so the agents you trace can also use their history. Installed once into a repo, they're discoverable by Claude Code, Cursor, Copilot, and any other client that reads the .agents/skills/ or .claude/skills/ conventions.

The bundled skills

use-thirdeye

Basic CLI fluency. Teaches an agent to:

• Enable tracing for a platform (thirdeye add).
• Search and retrieve sessions (thirdeye list, thirdeye events, thirdeye event, thirdeye tail, thirdeye search).
• Debug failed tool calls by filtering events and reading payloads.
• Read per-turn token usage from thirdeye usage.
• Tag and curate events (thirdeye tag, thirdeye tags).

Reach for this skill when you want your agent to introspect a past session — yours or its own — to debug, summarize, or follow up on prior work.

thirdeye-evals

The full eval workflow. Teaches an agent to:

• List and show shipped rubrics; create custom ones with thirdeye eval def create.
• Dispatch evaluators with thirdeye eval run (foreground or --background).
• Read results with thirdeye eval show / thirdeye eval list.
• Interpret per-turn findings anchored to event seq.
• Apply common patterns (token efficiency, tool quality, error recovery, task adherence, redundancy).

Pair it with use-thirdeye — the eval skill assumes you can already navigate sessions.

thirdeye-review

A workflow for auditing other agents' traces to find inefficiencies and propose fixes. Teaches an agent to:

• Scope a review cohort by agent type, repo, and time window.
• Cluster sessions by the task they were performing.
• Audit tool choices and execution patterns against known waste signatures (re-reads, bash-grep, retry loops, unbounded payloads).
• Quantify the token cost or error rate behind each pattern.
• Tie each finding back to its invocation surface — prompt, skill description, conventions file, plan task, or tool allowlist — and propose a concrete edit.

Use this when asked to reduce agent costs, improve results, or suggest new skills based on what you observe in past traces. Output is a structured report whose recommendations are each tied to a specific file or flag.

thirdeye-logfire

Setup and diagnostics for the optional Logfire export. Teaches an agent to:

• Check export health with thirdeye logfire status.
• Reinstall thirdeye if Logfire dependencies are missing, and enable export with a project write token.
• Connect the Logfire MCP server for querying exported traces.
• Troubleshoot missing spans and disable export cleanly.

It handles thirdeye's own export path only — it won't instrument an unrelated application with the Logfire SDK.

thirdeye-filter

A directive used by the browser UI's Ask panel to translate natural-language queries ("long-lasting claude runs this week") into the structured filter JSON the UI applies to your sessions or search view. Installed automatically with the other skills; not normally invoked directly by you or your agents.

Installing Skills from the Download

If you don't have thirdeye installed yet, you can bootstrap the process by downloading the skills and placing them manually.

1. Download the skills zip using the button above

2. Unzip into your project's .agents/skills/ directory:

mkdir -p .agents/skills
unzip thirdeye-skills.zip -d .agents/skills/

This creates:

.agents/skills/
├── use-thirdeye/
│   ├── SKILL.md
│   └── references/
├── thirdeye-evals/
│   ├── SKILL.md
│   └── references/
├── thirdeye-review/
│   ├── SKILL.md
│   └── references/
├── thirdeye-logfire/
│   ├── SKILL.md
│   └── references/
└── thirdeye-filter/
    └── SKILL.md

3. For Claude Code users, also copy to .claude/skills/:

mkdir -p .claude/skills
cp -r .agents/skills/* .claude/skills/

Unzipping anywhere

The zip has no fixed home — it expands to one directory per skill, so any folder your client scans for skills is a valid destination. Unzip it user-level to get the skills in every repo at once:

mkdir -p ~/.claude/skills
unzip thirdeye-skills.zip -d ~/.claude/skills/

Or pick a subset by unzipping only the paths you want:

unzip thirdeye-skills.zip 'use-thirdeye/*' 'thirdeye-evals/*' -d .agents/skills/

Copies from the zip are frozen at download time — they won't refresh when you upgrade thirdeye. Once thirdeye is installed, prefer thirdeye skills add, which symlinks instead so the content tracks your installed version.

Installing into a repo

thirdeye skills list                              # show bundled skill names
thirdeye skills add                               # all skills → .agents/skills/
thirdeye skills add --claude                      # → .claude/skills/
thirdeye skills add --codex                       # → .codex/skills/
thirdeye skills add --claude --codex              # both at once
thirdeye skills add --only thirdeye-review        # just one
thirdeye skills add --force                       # replace existing entries

With no target flag, thirdeye skills add writes into .agents/skills/ for cross-client discoverability. --claude and --codex select the client-specific conventions instead, and they can be combined to install into both in a single run.

Installing anywhere

-p/--path takes any directory — relative or absolute, inside the repo or outside it — and installs each skill as <dir>/<skill-name>. Parent directories are created for you.

thirdeye skills add -p ~/.claude/skills           # user-level, every repo
thirdeye skills add -p ~/.codex/skills            # user-level, Codex
thirdeye skills add -p /srv/shared/agent-skills   # a shared location
thirdeye skills add -p ../sibling-repo/.agents/skills

Installing to a user-level path makes the skills available in every project without a per-repo install, which is usually what you want if you use thirdeye across many repos:

thirdeye skills add -p ~/.claude/skills --force

-p/--path accepts a single folder and can't be combined with --claude/--codex — run it once per destination if you need several:

thirdeye skills add -p ~/.claude/skills
thirdeye skills add -p ~/.codex/skills

It composes with the other flags as expected:

thirdeye skills add -p ~/.claude/skills --only use-thirdeye --only thirdeye-evals

Symlinks, not copies

Skills install as symlinks pointing at the bundled skill content inside the installed thirdeye package. So when you upgrade thirdeye (brew upgrade thirdeye or pipx upgrade thrdi), the skill content refreshes automatically in every repo that has them installed — no per-repo reinstall required.

If you want a frozen copy that won't move with thirdeye upgrades, install once and then replace the symlink with cp -RL.

On Windows, creating a symlink requires Developer Mode or administrator rights. Without either, thirdeye skills add falls back to copying the skill directory instead. A copied skill doesn't track an upgrade the way a symlink does — after pipx upgrade thrdi (or uv tool upgrade thrdi), rerun thirdeye skills add --force to refresh the copies.

Using a skill

Claude Code

After install, Claude Code picks up skills under .agents/skills/ and .claude/skills/ automatically. Mention the skill by name in your prompt — for example, "use the use-thirdeye skill to find sessions that touched auth.py last week" — and Claude will load the skill instructions.

Cursor / Codex / Copilot / Gemini

Any client that reads the .agents/skills/ convention will discover the same files. For clients that don't, point them at the SKILL.md directly — the file is plain markdown and the rest of the skill (the references/ directory) is just supplementary content the agent reads when needed.

Tips

• Install them all together unless you specifically want a subset — agents tend to chain "find a problem" → "grade the problem" → "review the cohort" naturally once the surfaces are on hand.
• Re-run thirdeye skills add if you've moved or renamed your skills directory; it's idempotent and --force will replace stale entries.
• Audit what the skill does. Skills are markdown — read them. They're at .agents/skills/use-thirdeye/SKILL.md (or wherever you installed them) and link out to the reference files for each workflow.