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 phenomenoner/ai-agent-thinkroom --skill thinkroom-operategit clone --depth 1 https://github.com/phenomenoner/ai-agent-thinkroomWrote 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/phenomenoner/ai-agent-thinkroom/thinkroom-operate)<a href="https://agentmods.dev/skills/phenomenoner/ai-agent-thinkroom/thinkroom-operate"><img src="https://agentmods.dev/badge/skills/phenomenoner/ai-agent-thinkroom/thinkroom-operate/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/phenomenoner/ai-agent-thinkroom/thinkroom-operate"><img src="https://agentmods.dev/badge/skills/phenomenoner/ai-agent-thinkroom/thinkroom-operate.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 | $0.00010 | $0.00932 |
| Opus 5 | $0.00005 | $0.00466 |
| Sonnet 5 | $0.00002 | $0.00186 |
| Haiku 4.5 | $0.00001 | $0.00093 |
Grade A, and why
thinkroom-operate 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 4d 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 — 101 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Thinkroom operate
Use this skill after thinkroom-trigger passes, or whenever a user asks to submit, poll, inspect, cancel, or interpret a Thinkroom job.
Before submission
- Verify the service is ready on its configured loopback endpoint.
- Frame one explicit question from 10 through 10,000 characters and shared context no longer than 100,000 characters.
- Select one supported domain:
generic,coding, ortrading. For product design, organizational strategy, policy, due diligence, or general research synthesis, omit the field and use thegenericdefault. Usecodingonly when the branches must reason about software or source code, andtradingonly for market/trading decision support. Never send a descriptive label such asproduct_design,architecture, orstrategy. - Use an integer branch count from 2 through 6, choosing the smallest count that covers distinct perspectives.
- Use an idempotency key whenever a transport retry or duplicate submission is possible. It must be 1 through 128 printable non-space ASCII characters.
- For MCP, send only fields exposed by the callable tool schema. In particular, do not invent a
deadline_secondsfield when the current MCP tool does not expose one.
Submit
CLI:
thinkroom research \
--question "Should we adopt this design?" \
--idempotency-key <stable-key>
MCP: call thinkroom_research with question, optional context, domain, branch_count, and idempotency_key.
Submission returns a job handle. It does not imply the research succeeded.
If MCP returns INVALID_ARGUMENT without field-level details, first compare the exact submitted
payload with the strict fields above and the callable tool schema. The most common correction for a
natural-language research category is to remove domain and accept generic. Make at most one
retry only when that comparison produces a concrete schema correction. Preserve the question,
context, branch count, and exact idempotency key so the correction cannot silently become a new
research request. If the submitted payload was already schema-correct, do not retry unchanged;
report the opaque rejection and inspect the service boundary. Do not poll or cancel until a
submission returns a job handle.
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 4d ago Changed · +10 lines 045ef7c11582
- 9d ago First seen · 91 lines · 10 tokens per session scan A 4cc408d77e97
thinkroom-operate is a skill published in the GitHub repository phenomenoner/ai-agent-thinkroom (2 stars, last pushed 4d ago), licensed MIT. It adds 10 tokens to every session and 932 once invoked, about $0.0001 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
hunt-fastapi
Hunt FastAPI-specific vulnerabilities: dependency injection gaps, Pydantic coercion, and OpenAPI mining.
research-intelligence
Use when a Hermes agent must perform public-source research, source scouting, evidence grading, competitor/tool comparison, community-signal analysis, or decision-ready brief writing without private data or credentials.
youtube-research-pack
Use when a Hermes researcher must search public YouTube, inspect videos/channels/playlists, collect subtitles, analyze public signals, or run a bounded YouTube intelligence workflow without cookies or private data.
markitdown-document-ingestion
Convert public research documents and mixed file formats into Markdown before evidence review. Use for PDF, DOCX, PPTX, XLSX, HTML, CSV/JSON/XML, EPUB, ZIP bundles, and document intake before summaries, source ledgers, or research briefs.
codeflow-adoption-router
Route the current job to the right Codeflow surface without overclaiming hosted, write-capable MCP, or published plugin support.
link-curator-dashboard
Maintain and debug the link-curator web dashboard (port 8090). Separate process from the official Hermes dashboard.