CtrlK
BlogDocsLog inGet started
Tessl Logo

talk-douglas-training-ai-on-your-own-code

Answers questions about Brian Douglas's talk on training AI on your own code. Use when a user asks about Brian Douglas's pipeline for capturing agent sessions, extracting skills from traces, fine-tuning small local models, tapes/steros tooling, SFT vs DPO decisions, or wants to apply his agent telemetry and training data approach to their own work with Claude Code, QLoRA, or parallel agents.

72

Quality

88%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

SecuritybySnyk

Passed

No findings from the security scan

SKILL.md
Quality
Evals
Security

Quality

Content

82%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 clean, efficient instruction-only reference skill with a clear grounding workflow and safety rules. It could be strengthened by making the example response fully concrete and confirming the referenced bundle files exist.

Suggestions

Replace the '[safe excerpts from transcript.md]' placeholder in the Good response example with a short real quoted excerpt to model the expected citation behavior concretely.

Ensure outline.md, transcript.md, and quote.md are present in the skill bundle so the referenced grounding workflow is actually executable.

Add an explicit validate-fix-retry loop (e.g., if a claim cannot be sourced, state it is uncovered rather than searching again) to raise the workflow checkpoint quality.

DimensionReasoningScore

Conciseness

The body is lean and assumes Claude's competence; it explains no general concepts and every section (grounding rules, response format, quotes pointer, safety) earns its place.

5 / 5

Actionability

Concrete, specific guidance is given ('read outline.md to locate the relevant section, then read that section of transcript.md', 'check quote.md first'), but the 'Good response' example uses a placeholder '[safe excerpts from transcript.md]' rather than a fully concrete model answer.

4 / 5

Workflow Clarity

The numbered grounding rules give a clear sequence with a validation checkpoint ('If a claim isn't in transcript.md, say so explicitly'), but the error-recovery feedback loop is lighter than the anchor-5 example's validate-fix-retry pattern.

4 / 5

Progressive Disclosure

The overview body points to clearly-signaled one-level-deep references (outline.md, transcript.md, quote.md), but the referenced files were not found in the bundle directories, so their existence cannot be confirmed.

4 / 5

Total

17

/

20

Passed

Description

95%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 well-constructed reference-skill description with explicit what/when structure and strong natural trigger terms. Minor room to sharpen the 'what' as a set of distinct actions rather than topical coverage.

DimensionReasoningScore

Specificity

The description names the domain and multiple concrete coverage areas ('capturing agent sessions, extracting skills from traces, fine-tuning small local models, tapes/steros tooling, SFT vs DPO decisions'), but these are topics rather than a comprehensive list of distinct actions, so it sits just below a 5.

4 / 5

Completeness

It explicitly answers 'what' ('Answers questions about Brian Douglas's talk…') and 'when' ('Use when a user asks about…') with concrete trigger phrases.

5 / 5

Trigger Term Quality

It includes a rich set of natural phrases users would say ('training AI on your own code', 'fine-tuning small local models', 'SFT vs DPO', 'Claude Code', 'QLoRA', 'parallel agents'), covering synonyms and tool names comprehensively.

5 / 5

Distinctiveness Conflict Risk

It targets a single named talk by a specific speaker with distinct triggers, so it is unlikely to fire for unrelated skills.

5 / 5

Total

19

/

20

Passed

Validation

93%

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

Validation15 / 16 Passed

Validation for skill structure

CriteriaDescriptionResult

metadata_version

'metadata.version' is missing

Warning

Total

15

/

16

Passed

Repository
jscraik/Agent-Skills
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.