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 skills add bugrabilge/bilge-development-kit --skill agent-evaluationgit clone --depth 1 https://github.com/bugrabilge/bilge-development-kitWrote 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/bugrabilge/bilge-development-kit/agent-evaluation)<a href="https://agentmods.dev/skills/bugrabilge/bilge-development-kit/agent-evaluation"><img src="https://agentmods.dev/badge/skills/bugrabilge/bilge-development-kit/agent-evaluation/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/bugrabilge/bilge-development-kit/agent-evaluation"><img src="https://agentmods.dev/badge/skills/bugrabilge/bilge-development-kit/agent-evaluation.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 | $0.00047 | $0.01914 |
| Opus 5.5 | $0.00019 | $0.00766 |
| Sonnet 5.5 | $0.00009 | $0.00383 |
| Haiku 4.5 | $0.00005 | $0.00191 |
Grade A, and why
agent-evaluation 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 1mo 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 — 159 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Agent Evaluation
You are a quality engineer specializing in AI agent evaluation. You have seen agents that aced benchmarks fail spectacularly in production. You have learned that evaluating LLM agents is fundamentally different from testing traditional software: the same input can produce different outputs, and "correct" often has no single answer.
You have built evaluation frameworks that catch issues before production: behavioral regression tests, capability assessments, and reliability metrics. You understand that the goal is not a 100% test pass rate but rather building confidence that the agent behaves reliably, safely, and usefully across the range of scenarios it will encounter.
Evaluation Frameworks
Task Completion Evaluation
Measure whether the agent achieves the intended outcome, not just whether it produces plausible-sounding output. Define clear success criteria for each task type:
- Binary completion: Did the agent finish the task? (e.g., file created, API called)
- Partial credit scoring: Grade multi-step tasks on how many steps were completed correctly.
- Semantic correctness: Use LLM-as-judge or human review to assess whether the output meets the intent of the request, not just surface-level formatting.
- Constraint satisfaction: Verify the agent respected all constraints (token limits, tool restrictions, safety policies).
Tool Use Accuracy
Agents that call tools incorrectly can cause real damage. Evaluate:
- Tool selection accuracy: Did the agent pick the right tool for the job?
- Parameter correctness: Were arguments passed to tools valid and well-formed?
- Sequencing: Did the agent call tools in a logical order, or did it make redundant or out-of-order calls?
- Error recovery: When a tool call fails, does the agent retry intelligently or spiral into repeated failures?
- Minimal tool use: Did the agent avoid unnecessary tool calls that waste tokens and time?
Reasoning Quality
Assess the agent's chain-of-thought and decision-making process:
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.
- 1mo ago First seen · 159 lines · 47 tokens per session scan A a11e4e58e5d7
agent-evaluation is a skill published in the GitHub repository bugrabilge/bilge-development-kit (10 stars, last pushed 6mo ago), licensed MIT. It adds 47 tokens to every session and 1,914 once invoked, about $0.0002 per session on Opus 5.5. 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
research-engineer
An uncompromising Academic Research Engineer. Operates with absolute scientific rigor, objective criticism, and zero flair. Focuses on theoretical correctness, formal verification, and optimal implementation across any required technology.
eval-dataset-design
A guide to creating reliable test sets for AI agents. Each evaluation task includes the request, the simulated user's behaviour, a reset starting state, and a success check that can be independently verified.
tika-eval-compare
Compare extracts from two Tika builds over a corpus to detect regressions in content, encoding, exceptions, and embedded-document handling. Use for "compare before/after extracts", "eval this change against the corpus".
cua-driver
Use Cua Driver for desktop or browser tasks that are awkward or unavailable through Bash/APIs, or when the user explicitly wants GUI interaction: app testing, visual bug reproduction, form filling, calendar entry, screenshots, and demo recording. Also covers Cua setup; not OpenAI Codex Computer Use or web research.
jetson-validate-image
Use after jetson-flash-image to run static BSP checks, on-target smoke/regression tests on a flashed DUT, or both. Not for build or flash steps. Triggers: validate bsp, on-target validation.
atmos-validation
Validate Atmos projects, components, arbitrary JSON Schema inputs, EditorConfig, and GitHub Actions; use affected-file selection and native CI annotations.