Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/rctruta/sql-benchmarks-dagster/read-experiment-resultsnpx skills add rctruta/sql-benchmarks-dagster --skill read-experiment-resultsgit clone --depth 1 https://github.com/rctruta/sql-benchmarks-dagsterWrote 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/rctruta/sql-benchmarks-dagster/read-experiment-results)<a href="https://agentmods.dev/skills/rctruta/sql-benchmarks-dagster/read-experiment-results"><img src="https://agentmods.dev/badge/skills/rctruta/sql-benchmarks-dagster/read-experiment-results.svg" alt="Measured on agentmods" 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 | $0.00039 | $0.00724 |
| Opus 5 | $0.00019 | $0.00362 |
| Sonnet 5 | $0.00008 | $0.00145 |
| Haiku 4.5 | $0.00004 | $0.00072 |
Grade A, and why
read-experiment-results 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 5d 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 — 45 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Skill: read experiment results — pick the right tool for the question
Use when: an experiment's status has gone complete and you need to reason from the results.
Decision table
| Question shape | Tool | Why |
|---|---|---|
| First read of any completed experiment | python scripts/get_experiment_summary.py |
Compact: config identity + means + scaling ratios + prose narrative. Small payload, safe under tight context budgets. |
| "Who was faster on partition X?" | python scripts/get_means_by_partition.py |
Just means + sample counts. Cheapest projection when the question is per-partition speed, not ranking. |
| "How does X scale from small to large?" | python scripts/get_scaling_factor.py |
Returns ratios directly — adjacent_ratios, overall_ratio. Spares in-context arithmetic (which small models get wrong). |
| "Can I trust these numbers? / how noisy?" | python scripts/get_replication_stability.py |
std, coefficient of variation, min, max over raw per-replication durations. If sample_count=1 and std=0, the fragment predates raw capture — stability not measurable. |
| "Who won overall (across all partitions)?" | python scripts/compare_engines.py |
Ranked, aggregated across partitions. Flattens the scaling curve — don't use for scaling questions. |
| "Per-partition ranking with speedups" | python scripts/compare_engines_by_partition.py |
One ranking per partition. Use when you need speedup ratios AND the per-partition breakdown. |
| "I need every fragment's raw numbers" | python scripts/get_experiment_result.py |
The full payload. Use ONLY if the projections above don't answer the question. Can be many KB of JSON. |
Ordering caveat for get_scaling_factor
Partitions are sorted alphabetically, not semantically. ["small", "medium", "large"] becomes ["large", "medium", "small"]. The response includes partitions_order — check it and reinterpret if the semantic direction is reversed. overall_ratio < 1 means partitions get faster along the alpha order.
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.
- 5d ago First seen · 45 lines · 39 tokens per session scan A 745646ede177
read-experiment-results is a skill published in the GitHub repository rctruta/sql-benchmarks-dagster (2 stars, last pushed 1mo ago), licensed Apache-2.0. It adds 39 tokens to every session and 724 once invoked, about $0.0002 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
graphjin-eval
Create, extend, run, baseline, and diagnose GraphJin agent evaluations through the graphjin eval CLI.
graphjin-env-workflows
Use when running GraphJin's own environment and evaluation workflows in this repository — generating a suite, cloning or minting a world, authoring tasks, serving graded episodes, sampling, exporting trajectories, publishing a benchmark run — or when changing code those workflows depend on.
graphjin-env
Use when setting up a training or evaluation loop against a GraphJin agent environment — running the container, reading /health, driving episodes hosted or step-by-step or with your own agent over MCP, splitting train from eval, exporting trajectories, and deciding whether two rewards can be compared.
add-graphjin-database
Use when adding a new GraphJin database, warehouse, or CQL/NoSQL backend; building a simulator because no live service is available; wiring a dialect, discovery, tests, scripts, README/CONFIG/FEATURES, or website database support surfaces.
network-data-analysis
Ad-hoc read-only SQL analysis over exported network data (Zeek logs, Suricata eve.json, generated reports) using DuckDB. Use when aggregating across a packet capture's sessions, correlating IDS alerts with connection metadata, or answering counting and grouping questions that a per-log view cannot.
database-patterns
Database design and migration patterns for Alembic migrations, schema design (SQL/NoSQL), and database versioning. Use when creating migrations, designing schemas, normalizing data, managing database versions, or handling schema drift.