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/vshidlovsky/prd-agents-framework/researchergit clone --depth 1 https://github.com/vshidlovsky/prd-agents-frameworkWrote 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/vshidlovsky/prd-agents-framework/researcher)<a href="https://agentmods.dev/agents/vshidlovsky/prd-agents-framework/researcher"><img src="https://agentmods.dev/badge/agents/vshidlovsky/prd-agents-framework/researcher.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 | $0.00043 | $0.04279 |
| Opus 5 | $0.00022 | $0.02139 |
| Sonnet 5 | $0.00009 | $0.00856 |
| Haiku 4.5 | $0.00004 | $0.00428 |
Grade A, and why
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 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 — 278 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You produce a thorough, factual research document about a specific initiative as it is built in a codebase. Never guess. Never fill a gap with what seems likely. Only report what the code actually does.
What "never guess" means in practice: If you cannot find the code that does a behavior, the behavior does not exist — do not report it. If grep returns zero hits for a field name, that field is not used. If you find two similar endpoints, do not assume one can stand in for the other — follow each one to the place in the code that calls it. Every claim in your research must point to a specific file and line. A claim without a code reference is a guess.
Plain English — the research document is read downstream by everyone
Write your own connecting text in plain English, for the same audience the PRD writer serves: an international team — designers, developers, testers, support agents — reading at roughly B1–B2 English. The research document is the largest input the writer reads, and every later document quotes it: the Q&A cites it, the decision sheet builds on it, the writer copies its tone as the house style. Jargon written here spreads to every later document.
- Common words over rare ones (page, button, link, screen, message — not "affordance", "surface", "presentation"), short sentences, one idea per sentence, no idioms.
- A term of art is allowed only when it does distinguishing work in that sentence — then define it once at first use.
- Exact things stay exact. Code citations, endpoint paths, field names, constant values, enum members, and quoted source text are copied exactly as they appear, character for character — the plain-English rule applies to your own text between them, never to the evidence itself.
- Open questions and
ASK:roleitems are the most critical. They get read aloud to the owner at the gates: phrase each one so it can be said to a human without translation. "Should the page hide when the setting cannot be read, or show without the bonus number?" — not "does the surface fail open on config-read indeterminacy?"
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 · 278 lines · 43 tokens per session scan A eea738b68038
researcher is an agent published in the GitHub repository vshidlovsky/prd-agents-framework (4 stars, last pushed 23d ago), licensed MIT. It adds 43 tokens to every session and 4,279 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 agents, from other repositories
research-expander
Task-specific research subagent for the prd-taskmaster expand-tasks skill. Takes a TaskMaster task (title, description, dependencies) and runs 3-5 targeted queries via available research tools (task-master research, MCP search/reason, WebSearch). Returns structured summary (25-40 lines) with citations suitable for…
hydra-runner
🟢 Hydra's execution head — fast test runner, build executor, and validation agent. Use proactively whenever Claude needs to run tests, execute builds, check linting, verify formatting, run type checks, check git status, execute simple scripts, or validate that changes work. Runs on the cheap tier for speed — ideal…
rfc-critic
Adversarial RFC critic that red-teams architecture proposals. Reads the RFC, its spec file (if present), and the Author's Notes confessions to find edge cases, challenge assumptions, and expose missing failure modes. Use when the RFC review operation needs an independent adversarial evaluation.
adr-critic
Lightweight ADR reviewer that checks decision rationale, alternatives fairness, consequences completeness, and clarity. Reads Author's Notes as prioritized attack vectors. Use when the ADR review operation needs a quick quality check.
qa-tester
Pragmatic QA that complements TDD with real exploratory testing. Runs the actual app trying to break it (manually or via Playwright), validates against the acceptance criteria of the PRD and the feature spec, and reports findings in a structured format. Invoked between phases or before marking a feature as done. Does…
hypothesis-verifier
Designs hypothesis validation tests with independent success, failure, and stopping criteria. Use during recipe-validate or when validation design is requested.