Borrowing it
Nothing to install: this file belongs to sagar-shirwalkar/collibra-atlas. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/sagar-shirwalkar/collibra-atlas/main/.agents/skills/eval-driven-dev/SKILL.mdgit clone --depth 1 https://github.com/sagar-shirwalkar/collibra-atlasWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/sagar-shirwalkar/collibra-atlas/eval-driven-dev)<a href="https://agentmods.dev/skills/sagar-shirwalkar/collibra-atlas/eval-driven-dev"><img src="https://agentmods.dev/badge/skills/sagar-shirwalkar/collibra-atlas/eval-driven-dev/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/sagar-shirwalkar/collibra-atlas/eval-driven-dev"><img src="https://agentmods.dev/badge/skills/sagar-shirwalkar/collibra-atlas/eval-driven-dev.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00065 | $0.00823 |
| Opus 5 | $0.00032 | $0.00411 |
| Sonnet 5 | $0.00013 | $0.00165 |
| Haiku 4.5 | $0.00006 | $0.00082 |
Grade A, and why
eval-driven-dev scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 8d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 66 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Set up automated evaluation pipelines for Python LLM applications using pixie-qa. The eval exercises the real application code (routing, prompt assembly, LLM calls, response formatting) with controlled input data — nothing is mocked or stubbed except the external data sources the app reads.
Rule: The app's LLM calls must go to a real LLM. Do not replace, mock, stub, or intercept the LLM. Replacing the LLM makes the evaluation tautological.
Leading words
- Instrument — Add
wrap()calls at the app's data boundaries so the eval harness can inject controlled inputs and capture outputs. Makes the app testable without changing its logic. - Golden dataset — A curated set of inputs with expected behavior characteristics, covering the app's capabilities and known failure modes.
- Score — An evaluator's numeric judgment of an output (LLM-as-judge, similarity, deterministic check). Scores produce pass/fail decisions.
Phases
PHASE 1: Understand and define
Analyze the project, identify its entry points, and define eval criteria derived from real failure modes. Produces three artifacts in pixie_qa/: project analysis, entry point, and eval criteria.
PHASE 2: Instrument and reference trace
Add wrap() calls at data boundaries, implement a Runnable class that invokes the real entry point, and capture a reference trace that proves instrumentation works.
PHASE 3: Define evaluators
Map each eval criterion to a scoring function — built-in pixie evaluators, LLM-as-judge agent evaluators, or manual custom functions for deterministic checks.
PHASE 4: Build dataset
Create JSON entries that tie together the Runnable, evaluators, and use cases. Each entry specifies inputs, expected data, and which evaluators to apply.
PHASE 5: Run tests
Execute pixie test and fix mechanical issues. Once tests produce real evaluator scores, proceed to analysis.
PHASE 6: Analyze outcomes
Complete pending evaluations, analyze per-dataset and per-run results, and produce a prioritized action plan grounded in concrete data.
What ships with it
18 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
- references/1-a-project-analysis.md 5.7 KB
- references/1-b-entry-point.md 2.3 KB
- references/1-c-eval-criteria.md 7.5 KB
- references/2a-instrumentation.md 7.3 KB
- references/2b-implement-runnable.md 7.4 KB
- references/2c-capture-and-verify-trace.md 5.7 KB
- references/3-define-evaluators.md 10 KB
- references/4-build-dataset.md 22 KB
- references/5-run-tests.md 6.0 KB
- references/6-analyze-outcomes.md 16 KB
- references/evaluators.md 18 KB
- references/runnable-examples/cli-app.md 2.0 KB
- references/runnable-examples/fastapi-web-server.md 4.1 KB
- references/runnable-examples/standalone-function.md 1.8 KB
- references/testing-api.md 16 KB
- references/wrap-api.md 8.8 KB
- resources/setup.sh 3.1 KB runs code
- resources/verify_step6_completion.py 4.4 KB runs code
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 8d ago First seen · 66 lines · 65 tokens per session scan A fda936f019cf
eval-driven-dev is a skill published in the GitHub repository sagar-shirwalkar/collibra-atlas (0 stars, last pushed 1mo ago), licensed Apache-2.0. It adds 65 tokens to every session and 823 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
cli-eval
Create and run evaluation suites, watch live benchmark progress, view scorecards, compare model performance, and integrate eval runs with CI workflows from the CLI.
model-merging
Merge multiple fine-tuned models using mergekit to combine capabilities without retraining. Use when creating specialized models by blending domain-specific expertise (math + coding + chat), improving performance beyond single models, or experimenting rapidly with model variants. Covers SLERP, TIES-Merging, DARE, Task…
darwinian-evolver
Evolve prompts/regex/SQL/code with Imbue's evolution loop.
validate
Validate Semantica pipelines, extraction quality, graph schemas, and ontology consistency. Returns structured error/warning checklists. Uses PipelineValidator, PipelineBuilder.validatepipeline(), GraphValidator, and OntologyValidator. Sub-commands: pipeline, step, dependencies, extraction, graph, ontology, performance.
launching-evals
Run, monitor, analyze, and debug LLM evaluations via nemo-evaluator-launcher. Covers running evaluations, checking status and live progress, debugging failed runs, exporting artifacts and logs, and analyzing results. ALWAYS triggers on mentions of running evaluations, checking progress, debugging failed evals…
nemo-automodel-recipe-development
Create and modify NeMo AutoModel training and evaluation recipes, including YAML structure, builders, and execution flow.