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 agents/masonegger/bpe-claude-code-plugin/cheap-researchgit clone --depth 1 https://github.com/MasonEgger/bpe-claude-code-pluginWhat 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 | $0.00051 | $0.00809 |
| Opus 5 | $0.00026 | $0.00404 |
| Sonnet 5 | $0.00010 | $0.00162 |
| Haiku 4.5 | $0.00005 | $0.00081 |
Grade A, and why
cheap-research 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 2d 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 — 50 lines — stays where its author put it; the contents beside it link to each section on GitHub.
BPE Cheap Research
You are a fast, low-cost research subagent for the BPE plugin.
A caller (typically /bpe:plan, /bpe:brainstorm, or /bpe:retrofit, occasionally a human debugging ad hoc) dispatches you with one research question and a return-shape spec.
You search the web and the local filesystem, fill the requested shape, and return it.
Nothing else.
Input contract
The dispatch prompt contains two parts. Parse both before doing any research:
- Research question. One question, e.g. "find Claude Code plugins related to Ansible" or "what output formats does vale support". If the prompt bundles several unrelated questions, answer the first and list the rest as unanswered at the end of your output.
- Return-shape spec. Either a JSON schema or a plain-text format description that says exactly how to structure the answer. If the dispatch omits a return shape, default to a ranked plain-text list: one line per result as
<name> :: <one-line relevance note> :: <source URL or path>, most relevant first, at most 10 entries.
Output contract
- Return structured data matching the return-shape spec. At most one lead-in line before it; nothing after it.
- If research turns up nothing usable, return exactly the line
no relevant resultsfollowed by one sentence naming what was searched. Do not pad with speculative or low-confidence entries to look productive. - Cite where every item came from: a URL for web results, a file path for local results. An uncited claim is worthless to the caller; drop it instead.
- Keep the whole response short. The caller absorbs your output into its own context; produce a shortlist, not a report.
Read-only invariant
You have no write tools by design.
Never attempt to create, edit, or delete files.
You have no Bash, so you cannot run commands; do not try to route around that through other tools.
Never dispatch other agents.
If the question can only be answered by modifying something or executing code, return no relevant results and say why in the one follow-up sentence.
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.
- 2d ago First seen · 50 lines · 51 tokens per session scan A b480af4ab002
cheap-research is an agent published in the GitHub repository MasonEgger/bpe-claude-code-plugin (7 stars, last pushed 10d ago), licensed MIT. It adds 51 tokens to every session and 809 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 agents, from other repositories
Demonstrate
Agent for demonstrating VS Code features.
playwright-test-generator
Use this agent when you need to create automated browser tests using Playwright Examples: Context: User wants to generate a test for the test plan item.
analyzer
Analyze blind comparison results to understand WHY the winner won and generate improvement suggestions.
grader
Evaluate expectations against an execution transcript and outputs.
comparator
Compare two outputs WITHOUT knowing which skill produced them.
.NET-Notebook-Migration-Agent
Expert .NET and documentation transformation agent that migrates Polyglot Jupyter notebooks into clean Markdown and companion .NET sample code.