phoenix-cli

Debug LLM applications using the Phoenix CLI. Fetch traces, analyze errors, structure trace review with open coding and axial coding, inspect datasets, review experiments, query annotation configs, and use the GraphQL API. Use whenever the user is analyzing traces or spans, investigating LLM/agent failures, deciding what to do after instrumenting an app, building failure taxonomies, choosing what evals to write, or asking "what's going wrong", "what kinds of mistakes", or "where do I focus" — even without naming a technique.

Quality

87%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

High

Do not use without reviewing

Quality

Content

75%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

A dense, highly actionable CLI reference with comprehensive executable examples and well-gated destructive operations, supported by two clearly signaled one-level reference files. Its main weakness is conciseness: large filter-expression and JSON-shape reference blocks are inlined in SKILL.md rather than split into references, leaving room to tighten the body.

Suggestions

Move the three filter-expression sections (span/trace/session DSLs and their vocabulary enumerations) into a dedicated reference file, keeping only a short pointer and one example per grain in SKILL.md to reduce the body's token footprint.

Extract the full Trace/Span/Session JSON-shape blocks into a reference file, summarizing only the most-used fields inline.

Render the open-coding -> axial-coding -> evals workflow as an explicit numbered checklist with a verification checkpoint per step to lift workflow clarity toward 5.

Dimension	Reasoning	Score
Conciseness	Prose is high-signal and avoids explaining concepts Claude already knows, but ~620 lines inline large reference blocks (three filter-expression DSL sections with vocabulary enumerations, full JSON shapes, a 50-command invocation list) that could be tightened or split out, fitting the score-3 'mostly efficient but could be tightened' anchor.	3 / 5
Actionability	Pervasive copy-paste-ready commands with real flags, jq pipelines, exit codes, and concrete GraphQL queries (e.g. 'px trace list --last-n-minutes 60 --limit 20 --format raw --no-progress \| jq '.[] \| select(.status == "ERROR")'') covering common cases comprehensively.	5 / 5
Workflow Clarity	The open-coding -> axial-coding -> evals workflow is sequenced and destructive/batch operations carry explicit validation (env gate, confirmation, exit codes 6/3, tracesVerified/verification), so the destructive-cap does not apply; held at 4 because the top-level workflow is prose rather than an explicit per-step checklist with checkpoints (detail delegated to references).	4 / 5
Progressive Disclosure	Two real, one-level-deep reference files (open-coding.md, axial-coding.md) are clearly signaled via Quick Reference and Reference Categories tables with working relative paths; good structure with a minor gap that the bulky filter-DSL and JSON-shape material remains inlined rather than offloaded to references.	4 / 5
	Total	16 / 20 Passed

Description

100%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

A strong, third-person description that names a specific tool and many concrete capabilities, then pairs them with explicit, naturally voiced trigger phrases including colloquial ones. It clearly answers both what the skill does and when to use it with minimal conflict risk.

Dimension	Reasoning	Score
Specificity	Lists multiple concrete actions over a named domain — 'Fetch traces, analyze errors, structure trace review with open coding and axial coding, inspect datasets, review experiments, query annotation configs, and use the GraphQL API' — exceeding the score-5 anchor's breadth with no coverage gaps.	5 / 5
Completeness	Explicitly answers both 'what' (the Phoenix CLI action list) and 'when' ('Use whenever the user is analyzing traces or spans...') with concrete quoted trigger phrases, matching the score-5 anchor.	5 / 5
Trigger Term Quality	Includes natural user phrasings ('what's going wrong', 'what kinds of mistakes', 'where do I focus') plus synonyms like 'investigating LLM/agent failures' and 'building failure taxonomies', with an explicit catch-all ('even without naming a technique').	5 / 5
Distinctiveness Conflict Risk	Carves a clear LLM-observability debugging niche (traces/spans, instrumenting, failure taxonomies, evals) with distinct triggers and minimal overlap risk with unrelated skills.	5 / 5
	Total	20 / 20 Passed

Validation

87%

Warnings & errors only

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 14 / 16 Passed

Validation for skill structure

Criteria	Description	Result
skill_md_line_count	SKILL.md is long (633 lines); consider splitting into references/ and linking	Warning
referenced_paths_exist	Referenced path issues: 4 missing	Warning

	Total	14 / 16 Passed

Repository: Arize-ai/phoenix
Path: .agents/skills/phoenix-cli/SKILL.md
Commit: 8adf337

Reviewed: 1 day ago

Table of Contents

Discovery Implementation Validation

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.