CtrlK
BlogDocsLog inGet started
Tessl Logo

migrate

Migrate an application with hardcoded LLM prompts to a full LaunchDarkly AgentControl implementation in five stages: audit the code, wrap the call, move the tools, add tracking, attach evaluators. Use when the user wants to externalize model/prompt configuration, move from direct provider calls (OpenAI, Anthropic, Bedrock, Gemini, Strands) to a managed config, or stage a full hardcoded-to-LaunchDarkly migration.

67

Quality

81%

Does it follow best practices?

Impact

No eval scenarios have been run

SecuritybySnyk

Advisory

Suggest reviewing before use

SKILL.md
Quality
Evals
Security

Quality

Content

62%

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

This is a highly actionable and well-structured migration guide with excellent workflow clarity, explicit checkpoints, and concrete code examples for multiple languages and frameworks. However, it suffers significantly from verbosity — the same warnings are repeated across multiple sections, extensive inline detail that belongs in reference files bloats the main document, and explanations of concepts Claude already understands waste tokens. The progressive disclosure structure exists but is undermined by the sheer volume of content kept inline.

Suggestions

Move the coverage matrix table, the 'What NOT to Do' section, and the edge cases table into separate reference files (e.g., coverage-matrix.md, common-mistakes.md, edge-cases.md) and replace them with one-line references — this alone would cut the skill by ~40%.

Eliminate repeated warnings: the tracker-lifetime rules appear in the Stage 4 callout box, the Stage 4 workflow steps, and the 'What NOT to Do' section. State each rule once in the workflow step where it's actionable and remove the duplicates.

Remove explanatory commentary that Claude doesn't need (e.g., 'PDF stands for...'-level explanations like 'The set_config call is what initializes the singleton; .get() just returns it' or 'Each call mints a fresh runId that tags every event emitted from the turn so they can be correlated via exported events or downstream queries'). Trust Claude to understand SDK initialization patterns and event correlation.

DimensionReasoningScore

Conciseness

This skill is extremely verbose at ~700+ lines. It over-explains concepts Claude already knows (e.g., what offline mode is, how Python imports work, what AttributeError means), repeats the same warnings multiple times across sections (tracker lifetime rules appear in Stage 4, What NOT to Do, and the callout box), and includes extensive coverage matrices and edge case tables that could be in reference files. The three-failure-modes callout, the coverage table, and the 'What NOT to Do' section substantially duplicate guidance already given in the workflow steps.

1 / 3

Actionability

The skill provides fully executable, copy-paste-ready code examples for both Python and Node.js across all stages. Specific commands for package installation, concrete before/after code transformations, exact API calls with correct parameter names, and structured output templates make this highly actionable.

3 / 3

Workflow Clarity

The five-stage workflow is clearly sequenced with explicit validation checkpoints at each stage (sub-step 9 in Stage 2, sub-step 4 in Stage 3, sub-step 5 in Stage 4, sub-step 5 in Stage 5). The Stage 1 checkpoint with four explicit confirmation forms is a strong feedback loop. The hand-off model is clearly defined with explicit 'STOP' and 'Delegate' markers. Error recovery paths are specified (e.g., fallback verification, targeting flip requirement).

3 / 3

Progressive Disclosure

The skill references many external files (phase-1-analysis-checklist.md, before-after-examples.md, sdk-ai-tracker-patterns.md, agent-mode-frameworks.md, fallback-defaults-pattern.md, agent-graph-reference.md) which is good structure, but the SKILL.md itself is monolithic — it inlines enormous amounts of detail that should live in those reference files (full code examples for every tier, extensive edge case tables, repeated warnings). The coverage matrix table and the large 'What NOT to Do' section could be separate reference documents, keeping the main skill leaner.

2 / 3

Total

9

/

12

Passed

Description

100%

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

This is an excellent skill description that clearly articulates a specific migration workflow with five named stages, provides explicit trigger guidance via a 'Use when' clause, and includes natural keywords users would mention (provider names, 'hardcoded', 'externalize'). It is highly distinctive due to its focus on LaunchDarkly AgentControl and uses proper third-person voice throughout.

DimensionReasoningScore

Specificity

Lists five specific concrete stages (audit the code, wrap the call, move the tools, add tracking, attach evaluators) and names the overall action (migrate an application with hardcoded LLM prompts to LaunchDarkly AgentControl). Very specific and actionable.

3 / 3

Completeness

Clearly answers both 'what' (migrate hardcoded LLM prompts to LaunchDarkly AgentControl in five stages) and 'when' (explicit 'Use when' clause covering externalization, moving from direct provider calls, or staging a full migration).

3 / 3

Trigger Term Quality

Includes strong natural trigger terms: 'externalize model/prompt configuration', 'hardcoded', 'LaunchDarkly', 'migration', and lists specific provider names (OpenAI, Anthropic, Bedrock, Gemini, Strands) that users would naturally mention. Good coverage of variations.

3 / 3

Distinctiveness Conflict Risk

Highly distinctive with a clear niche: LaunchDarkly AgentControl migration specifically for LLM prompt management. The combination of LaunchDarkly, AgentControl, and the specific provider names makes it very unlikely to conflict with other skills.

3 / 3

Total

12

/

12

Passed

Validation

90%

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

Validation10 / 11 Passed

Validation for skill structure

CriteriaDescriptionResult

skill_md_line_count

SKILL.md is long (573 lines); consider splitting into references/ and linking

Warning

Total

10

/

11

Passed

Repository
launchdarkly/ai-tooling
Reviewed

Table of Contents

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.