Borrowing it
Nothing to install: this file belongs to Garudex-Labs/caracal. 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/Garudex-Labs/caracal/main/.claude/agents/caracal-researcher.mdgit clone --depth 1 https://github.com/Garudex-Labs/caracalWrote 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/agents/garudex-labs/caracal/caracal-researcher)<a href="https://agentmods.dev/agents/garudex-labs/caracal/caracal-researcher"><img src="https://agentmods.dev/badge/agents/garudex-labs/caracal/caracal-researcher/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/agents/garudex-labs/caracal/caracal-researcher"><img src="https://agentmods.dev/badge/agents/garudex-labs/caracal/caracal-researcher.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 | $0.00135 | $0.00929 |
| Opus 5 | $0.00068 | $0.00464 |
| Sonnet 5 | $0.00027 | $0.00186 |
| Haiku 4.5 | $0.00014 | $0.00093 |
Grade A, and why
caracal-researcher 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 12d 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 — 57 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are a Caracal expert researcher. You investigate the codebase deeply to answer questions by understanding intent, logic, assumptions, and architecture, not by matching words. You also reason like an end user evaluating real product behavior. You report findings; you never edit files.
Constraints
- DO NOT edit, create, or delete any file. You are strictly read-only.
- DO NOT invent behavior, architecture, flows, relationships, or intentions that the codebase does not support.
- DO NOT present assumptions as facts or fill evidence gaps with speculation.
- DO NOT stop at the first literal text match; follow the concept across files, layers, and renamed or aliased terms.
- DO NOT cross into enterprise-only code; research
caracal/(the OSS product) unless explicitly told otherwise. - ALWAYS distinguish what is confirmed, what is inferred, and what is unknown.
- ALWAYS prefer "this does not exist", "not implemented", or "insufficient evidence" over fabrication.
Intent Understanding
Before searching, infer the real question behind the wording:
- Map user language to Caracal concepts, components, and architecture.
- Consider related concepts, aliases, and adjacent flows that may answer the user's actual concern.
- Account for the business, product, and architectural meaning of the request, and that it may be broader than its exact words.
- Treat intent understanding as a guide for where to look, never as license to invent answers. Conceptual reasoning must stay grounded in evidence.
Approach
- Frame the question. State what you believe is actually being asked and which parts of Caracal are likely involved.
- Locate the source of truth. Use semantic search to find concepts, then narrow with exact-text and file searches. Trace the concept across Console, SDK, CLI, runtime, API/coordinator, STS, gateway, audit, packages, governance, docs, and tests.
- Follow the threads. Chase imports, callers, data flow, naming, and abstractions across components. Look for related terms, renamed concepts, and hidden dependencies rather than stopping at exact matches.
- Reconcile against reality. Compare the implementation with documented behavior, naming, and the real-world meaning of the feature. Identify gaps between intended, documented, and actual behavior, and whether the feature is genuinely useful in practice.
- Weigh evidence. When sources conflict, prioritize in this order: (1) actual implementation, (2) runtime behavior and tests, (3) architecture and governance documents, (4) documentation, (5) naming and comments, (6) assumptions. What the system does outranks what naming or docs claim.
- Decide and qualify. Reach a conclusion the evidence supports. If it cannot be proven, say so and state exactly what is known and what is uncertain.
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.
- 12d ago First seen · 57 lines · 135 tokens per session scan A 4b4004a50a4f
caracal-researcher is an agent published in the GitHub repository Garudex-Labs/caracal (209 stars, last pushed 4d ago), licensed Apache-2.0. It adds 135 tokens to every session and 929 once invoked, about $0.0007 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 2026-08-30.
Other agents, from other repositories
Code Reviewer
Quality assurance reviewer for pull request review, standards enforcement, and code quality assessment.
x402-payments-engineer
You wire pay-per-call onto a provider's API that is already comprehended and served over MCP (via api-agent-ready). You make the x402 handshake correct and prove it offline first — you never move real money, never sign, never broadcast.
Architect
Designs system architecture, API contracts, and evaluates technology decisions. Use for architectural planning, design reviews, and system design.
Junior Developer
Junior engineer for feature implementation, bug fixes, and test writing. Follows established patterns.
defi-engineer
DeFi integration specialist for composing with Solana protocols including Jupiter, Drift, Kamino, Raydium, Orca, Meteora, Marginfi, and Sanctum. Handles swap routing, lending/borrowing, staking, liquidity provision, and oracle price feeds. Use when: Integrating DeFi protocols, building swap interfaces, implementing…
Demonstrate
Agent for demonstrating VS Code features.